aboutsummaryrefslogtreecommitdiffstats
path: root/Documentation/filesystems/caching/cachefiles.txt
diff options
context:
space:
mode:
Diffstat (limited to 'Documentation/filesystems/caching/cachefiles.txt')
-rw-r--r--Documentation/filesystems/caching/cachefiles.txt501
1 files changed, 501 insertions, 0 deletions
diff --git a/Documentation/filesystems/caching/cachefiles.txt b/Documentation/filesystems/caching/cachefiles.txt
new file mode 100644
index 000000000000..c78a49b7bba6
--- /dev/null
+++ b/Documentation/filesystems/caching/cachefiles.txt
@@ -0,0 +1,501 @@
1 ===============================================
2 CacheFiles: CACHE ON ALREADY MOUNTED FILESYSTEM
3 ===============================================
4
5Contents:
6
7 (*) Overview.
8
9 (*) Requirements.
10
11 (*) Configuration.
12
13 (*) Starting the cache.
14
15 (*) Things to avoid.
16
17 (*) Cache culling.
18
19 (*) Cache structure.
20
21 (*) Security model and SELinux.
22
23 (*) A note on security.
24
25 (*) Statistical information.
26
27 (*) Debugging.
28
29
30========
31OVERVIEW
32========
33
34CacheFiles is a caching backend that's meant to use as a cache a directory on
35an already mounted filesystem of a local type (such as Ext3).
36
37CacheFiles uses a userspace daemon to do some of the cache management - such as
38reaping stale nodes and culling. This is called cachefilesd and lives in
39/sbin.
40
41The filesystem and data integrity of the cache are only as good as those of the
42filesystem providing the backing services. Note that CacheFiles does not
43attempt to journal anything since the journalling interfaces of the various
44filesystems are very specific in nature.
45
46CacheFiles creates a misc character device - "/dev/cachefiles" - that is used
47to communication with the daemon. Only one thing may have this open at once,
48and whilst it is open, a cache is at least partially in existence. The daemon
49opens this and sends commands down it to control the cache.
50
51CacheFiles is currently limited to a single cache.
52
53CacheFiles attempts to maintain at least a certain percentage of free space on
54the filesystem, shrinking the cache by culling the objects it contains to make
55space if necessary - see the "Cache Culling" section. This means it can be
56placed on the same medium as a live set of data, and will expand to make use of
57spare space and automatically contract when the set of data requires more
58space.
59
60
61============
62REQUIREMENTS
63============
64
65The use of CacheFiles and its daemon requires the following features to be
66available in the system and in the cache filesystem:
67
68 - dnotify.
69
70 - extended attributes (xattrs).
71
72 - openat() and friends.
73
74 - bmap() support on files in the filesystem (FIBMAP ioctl).
75
76 - The use of bmap() to detect a partial page at the end of the file.
77
78It is strongly recommended that the "dir_index" option is enabled on Ext3
79filesystems being used as a cache.
80
81
82=============
83CONFIGURATION
84=============
85
86The cache is configured by a script in /etc/cachefilesd.conf. These commands
87set up cache ready for use. The following script commands are available:
88
89 (*) brun <N>%
90 (*) bcull <N>%
91 (*) bstop <N>%
92 (*) frun <N>%
93 (*) fcull <N>%
94 (*) fstop <N>%
95
96 Configure the culling limits. Optional. See the section on culling
97 The defaults are 7% (run), 5% (cull) and 1% (stop) respectively.
98
99 The commands beginning with a 'b' are file space (block) limits, those
100 beginning with an 'f' are file count limits.
101
102 (*) dir <path>
103
104 Specify the directory containing the root of the cache. Mandatory.
105
106 (*) tag <name>
107
108 Specify a tag to FS-Cache to use in distinguishing multiple caches.
109 Optional. The default is "CacheFiles".
110
111 (*) debug <mask>
112
113 Specify a numeric bitmask to control debugging in the kernel module.
114 Optional. The default is zero (all off). The following values can be
115 OR'd into the mask to collect various information:
116
117 1 Turn on trace of function entry (_enter() macros)
118 2 Turn on trace of function exit (_leave() macros)
119 4 Turn on trace of internal debug points (_debug())
120
121 This mask can also be set through sysfs, eg:
122
123 echo 5 >/sys/modules/cachefiles/parameters/debug
124
125
126==================
127STARTING THE CACHE
128==================
129
130The cache is started by running the daemon. The daemon opens the cache device,
131configures the cache and tells it to begin caching. At that point the cache
132binds to fscache and the cache becomes live.
133
134The daemon is run as follows:
135
136 /sbin/cachefilesd [-d]* [-s] [-n] [-f <configfile>]
137
138The flags are:
139
140 (*) -d
141
142 Increase the debugging level. This can be specified multiple times and
143 is cumulative with itself.
144
145 (*) -s
146
147 Send messages to stderr instead of syslog.
148
149 (*) -n
150
151 Don't daemonise and go into background.
152
153 (*) -f <configfile>
154
155 Use an alternative configuration file rather than the default one.
156
157
158===============
159THINGS TO AVOID
160===============
161
162Do not mount other things within the cache as this will cause problems. The
163kernel module contains its own very cut-down path walking facility that ignores
164mountpoints, but the daemon can't avoid them.
165
166Do not create, rename or unlink files and directories in the cache whilst the
167cache is active, as this may cause the state to become uncertain.
168
169Renaming files in the cache might make objects appear to be other objects (the
170filename is part of the lookup key).
171
172Do not change or remove the extended attributes attached to cache files by the
173cache as this will cause the cache state management to get confused.
174
175Do not create files or directories in the cache, lest the cache get confused or
176serve incorrect data.
177
178Do not chmod files in the cache. The module creates things with minimal
179permissions to prevent random users being able to access them directly.
180
181
182=============
183CACHE CULLING
184=============
185
186The cache may need culling occasionally to make space. This involves
187discarding objects from the cache that have been used less recently than
188anything else. Culling is based on the access time of data objects. Empty
189directories are culled if not in use.
190
191Cache culling is done on the basis of the percentage of blocks and the
192percentage of files available in the underlying filesystem. There are six
193"limits":
194
195 (*) brun
196 (*) frun
197
198 If the amount of free space and the number of available files in the cache
199 rises above both these limits, then culling is turned off.
200
201 (*) bcull
202 (*) fcull
203
204 If the amount of available space or the number of available files in the
205 cache falls below either of these limits, then culling is started.
206
207 (*) bstop
208 (*) fstop
209
210 If the amount of available space or the number of available files in the
211 cache falls below either of these limits, then no further allocation of
212 disk space or files is permitted until culling has raised things above
213 these limits again.
214
215These must be configured thusly:
216
217 0 <= bstop < bcull < brun < 100
218 0 <= fstop < fcull < frun < 100
219
220Note that these are percentages of available space and available files, and do
221_not_ appear as 100 minus the percentage displayed by the "df" program.
222
223The userspace daemon scans the cache to build up a table of cullable objects.
224These are then culled in least recently used order. A new scan of the cache is
225started as soon as space is made in the table. Objects will be skipped if
226their atimes have changed or if the kernel module says it is still using them.
227
228
229===============
230CACHE STRUCTURE
231===============
232
233The CacheFiles module will create two directories in the directory it was
234given:
235
236 (*) cache/
237
238 (*) graveyard/
239
240The active cache objects all reside in the first directory. The CacheFiles
241kernel module moves any retired or culled objects that it can't simply unlink
242to the graveyard from which the daemon will actually delete them.
243
244The daemon uses dnotify to monitor the graveyard directory, and will delete
245anything that appears therein.
246
247
248The module represents index objects as directories with the filename "I..." or
249"J...". Note that the "cache/" directory is itself a special index.
250
251Data objects are represented as files if they have no children, or directories
252if they do. Their filenames all begin "D..." or "E...". If represented as a
253directory, data objects will have a file in the directory called "data" that
254actually holds the data.
255
256Special objects are similar to data objects, except their filenames begin
257"S..." or "T...".
258
259
260If an object has children, then it will be represented as a directory.
261Immediately in the representative directory are a collection of directories
262named for hash values of the child object keys with an '@' prepended. Into
263this directory, if possible, will be placed the representations of the child
264objects:
265
266 INDEX INDEX INDEX DATA FILES
267 ========= ========== ================================= ================
268 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400
269 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...DB1ry
270 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...N22ry
271 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...FP1ry
272
273
274If the key is so long that it exceeds NAME_MAX with the decorations added on to
275it, then it will be cut into pieces, the first few of which will be used to
276make a nest of directories, and the last one of which will be the objects
277inside the last directory. The names of the intermediate directories will have
278'+' prepended:
279
280 J1223/@23/+xy...z/+kl...m/Epqr
281
282
283Note that keys are raw data, and not only may they exceed NAME_MAX in size,
284they may also contain things like '/' and NUL characters, and so they may not
285be suitable for turning directly into a filename.
286
287To handle this, CacheFiles will use a suitably printable filename directly and
288"base-64" encode ones that aren't directly suitable. The two versions of
289object filenames indicate the encoding:
290
291 OBJECT TYPE PRINTABLE ENCODED
292 =============== =============== ===============
293 Index "I..." "J..."
294 Data "D..." "E..."
295 Special "S..." "T..."
296
297Intermediate directories are always "@" or "+" as appropriate.
298
299
300Each object in the cache has an extended attribute label that holds the object
301type ID (required to distinguish special objects) and the auxiliary data from
302the netfs. The latter is used to detect stale objects in the cache and update
303or retire them.
304
305
306Note that CacheFiles will erase from the cache any file it doesn't recognise or
307any file of an incorrect type (such as a FIFO file or a device file).
308
309
310==========================
311SECURITY MODEL AND SELINUX
312==========================
313
314CacheFiles is implemented to deal properly with the LSM security features of
315the Linux kernel and the SELinux facility.
316
317One of the problems that CacheFiles faces is that it is generally acting on
318behalf of a process, and running in that process's context, and that includes a
319security context that is not appropriate for accessing the cache - either
320because the files in the cache are inaccessible to that process, or because if
321the process creates a file in the cache, that file may be inaccessible to other
322processes.
323
324The way CacheFiles works is to temporarily change the security context (fsuid,
325fsgid and actor security label) that the process acts as - without changing the
326security context of the process when it the target of an operation performed by
327some other process (so signalling and suchlike still work correctly).
328
329
330When the CacheFiles module is asked to bind to its cache, it:
331
332 (1) Finds the security label attached to the root cache directory and uses
333 that as the security label with which it will create files. By default,
334 this is:
335
336 cachefiles_var_t
337
338 (2) Finds the security label of the process which issued the bind request
339 (presumed to be the cachefilesd daemon), which by default will be:
340
341 cachefilesd_t
342
343 and asks LSM to supply a security ID as which it should act given the
344 daemon's label. By default, this will be:
345
346 cachefiles_kernel_t
347
348 SELinux transitions the daemon's security ID to the module's security ID
349 based on a rule of this form in the policy.
350
351 type_transition <daemon's-ID> kernel_t : process <module's-ID>;
352
353 For instance:
354
355 type_transition cachefilesd_t kernel_t : process cachefiles_kernel_t;
356
357
358The module's security ID gives it permission to create, move and remove files
359and directories in the cache, to find and access directories and files in the
360cache, to set and access extended attributes on cache objects, and to read and
361write files in the cache.
362
363The daemon's security ID gives it only a very restricted set of permissions: it
364may scan directories, stat files and erase files and directories. It may
365not read or write files in the cache, and so it is precluded from accessing the
366data cached therein; nor is it permitted to create new files in the cache.
367
368
369There are policy source files available in:
370
371 http://people.redhat.com/~dhowells/fscache/cachefilesd-0.8.tar.bz2
372
373and later versions. In that tarball, see the files:
374
375 cachefilesd.te
376 cachefilesd.fc
377 cachefilesd.if
378
379They are built and installed directly by the RPM.
380
381If a non-RPM based system is being used, then copy the above files to their own
382directory and run:
383
384 make -f /usr/share/selinux/devel/Makefile
385 semodule -i cachefilesd.pp
386
387You will need checkpolicy and selinux-policy-devel installed prior to the
388build.
389
390
391By default, the cache is located in /var/fscache, but if it is desirable that
392it should be elsewhere, than either the above policy files must be altered, or
393an auxiliary policy must be installed to label the alternate location of the
394cache.
395
396For instructions on how to add an auxiliary policy to enable the cache to be
397located elsewhere when SELinux is in enforcing mode, please see:
398
399 /usr/share/doc/cachefilesd-*/move-cache.txt
400
401When the cachefilesd rpm is installed; alternatively, the document can be found
402in the sources.
403
404
405==================
406A NOTE ON SECURITY
407==================
408
409CacheFiles makes use of the split security in the task_struct. It allocates
410its own task_security structure, and redirects current->act_as to point to it
411when it acts on behalf of another process, in that process's context.
412
413The reason it does this is that it calls vfs_mkdir() and suchlike rather than
414bypassing security and calling inode ops directly. Therefore the VFS and LSM
415may deny the CacheFiles access to the cache data because under some
416circumstances the caching code is running in the security context of whatever
417process issued the original syscall on the netfs.
418
419Furthermore, should CacheFiles create a file or directory, the security
420parameters with that object is created (UID, GID, security label) would be
421derived from that process that issued the system call, thus potentially
422preventing other processes from accessing the cache - including CacheFiles's
423cache management daemon (cachefilesd).
424
425What is required is to temporarily override the security of the process that
426issued the system call. We can't, however, just do an in-place change of the
427security data as that affects the process as an object, not just as a subject.
428This means it may lose signals or ptrace events for example, and affects what
429the process looks like in /proc.
430
431So CacheFiles makes use of a logical split in the security between the
432objective security (task->sec) and the subjective security (task->act_as). The
433objective security holds the intrinsic security properties of a process and is
434never overridden. This is what appears in /proc, and is what is used when a
435process is the target of an operation by some other process (SIGKILL for
436example).
437
438The subjective security holds the active security properties of a process, and
439may be overridden. This is not seen externally, and is used whan a process
440acts upon another object, for example SIGKILLing another process or opening a
441file.
442
443LSM hooks exist that allow SELinux (or Smack or whatever) to reject a request
444for CacheFiles to run in a context of a specific security label, or to create
445files and directories with another security label.
446
447
448=======================
449STATISTICAL INFORMATION
450=======================
451
452If FS-Cache is compiled with the following option enabled:
453
454 CONFIG_CACHEFILES_HISTOGRAM=y
455
456then it will gather certain statistics and display them through a proc file.
457
458 (*) /proc/fs/cachefiles/histogram
459
460 cat /proc/fs/cachefiles/histogram
461 JIFS SECS LOOKUPS MKDIRS CREATES
462 ===== ===== ========= ========= =========
463
464 This shows the breakdown of the number of times each amount of time
465 between 0 jiffies and HZ-1 jiffies a variety of tasks took to run. The
466 columns are as follows:
467
468 COLUMN TIME MEASUREMENT
469 ======= =======================================================
470 LOOKUPS Length of time to perform a lookup on the backing fs
471 MKDIRS Length of time to perform a mkdir on the backing fs
472 CREATES Length of time to perform a create on the backing fs
473
474 Each row shows the number of events that took a particular range of times.
475 Each step is 1 jiffy in size. The JIFS column indicates the particular
476 jiffy range covered, and the SECS field the equivalent number of seconds.
477
478
479=========
480DEBUGGING
481=========
482
483If CONFIG_CACHEFILES_DEBUG is enabled, the CacheFiles facility can have runtime
484debugging enabled by adjusting the value in:
485
486 /sys/module/cachefiles/parameters/debug
487
488This is a bitmask of debugging streams to enable:
489
490 BIT VALUE STREAM POINT
491 ======= ======= =============================== =======================
492 0 1 General Function entry trace
493 1 2 Function exit trace
494 2 4 General
495
496The appropriate set of values should be OR'd together and the result written to
497the control file. For example:
498
499 echo $((1|4|8)) >/sys/module/cachefiles/parameters/debug
500
501will turn on all function entry debugging.