aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorDavid Howells <dhowells@redhat.com>2009-04-03 11:42:41 -0400
committerDavid Howells <dhowells@redhat.com>2009-04-03 11:42:41 -0400
commit9ae326a69004dea8af2dae4fde58de27db700a8d (patch)
tree3a1d88a6e297989bfbd17648b398c7aa5ef9bf30 /fs
parent800a964787faef3509d194fa33268628c3d1daa9 (diff)
CacheFiles: A cache that backs onto a mounted filesystem
Add an FS-Cache cache-backend that permits a mounted filesystem to be used as a backing store for the cache. CacheFiles uses a userspace daemon to do some of the cache management - such as reaping stale nodes and culling. This is called cachefilesd and lives in /sbin. The source for the daemon can be downloaded from: http://people.redhat.com/~dhowells/cachefs/cachefilesd.c And an example configuration from: http://people.redhat.com/~dhowells/cachefs/cachefilesd.conf The filesystem and data integrity of the cache are only as good as those of the filesystem providing the backing services. Note that CacheFiles does not attempt to journal anything since the journalling interfaces of the various filesystems are very specific in nature. CacheFiles creates a misc character device - "/dev/cachefiles" - that is used to communication with the daemon. Only one thing may have this open at once, and whilst it is open, a cache is at least partially in existence. The daemon opens this and sends commands down it to control the cache. CacheFiles is currently limited to a single cache. CacheFiles attempts to maintain at least a certain percentage of free space on the filesystem, shrinking the cache by culling the objects it contains to make space if necessary - see the "Cache Culling" section. This means it can be placed on the same medium as a live set of data, and will expand to make use of spare space and automatically contract when the set of data requires more space. ============ REQUIREMENTS ============ The use of CacheFiles and its daemon requires the following features to be available in the system and in the cache filesystem: - dnotify. - extended attributes (xattrs). - openat() and friends. - bmap() support on files in the filesystem (FIBMAP ioctl). - The use of bmap() to detect a partial page at the end of the file. It is strongly recommended that the "dir_index" option is enabled on Ext3 filesystems being used as a cache. ============= CONFIGURATION ============= The cache is configured by a script in /etc/cachefilesd.conf. These commands set up cache ready for use. The following script commands are available: (*) brun <N>% (*) bcull <N>% (*) bstop <N>% (*) frun <N>% (*) fcull <N>% (*) fstop <N>% Configure the culling limits. Optional. See the section on culling The defaults are 7% (run), 5% (cull) and 1% (stop) respectively. The commands beginning with a 'b' are file space (block) limits, those beginning with an 'f' are file count limits. (*) dir <path> Specify the directory containing the root of the cache. Mandatory. (*) tag <name> Specify a tag to FS-Cache to use in distinguishing multiple caches. Optional. The default is "CacheFiles". (*) debug <mask> Specify a numeric bitmask to control debugging in the kernel module. Optional. The default is zero (all off). The following values can be OR'd into the mask to collect various information: 1 Turn on trace of function entry (_enter() macros) 2 Turn on trace of function exit (_leave() macros) 4 Turn on trace of internal debug points (_debug()) This mask can also be set through sysfs, eg: echo 5 >/sys/modules/cachefiles/parameters/debug ================== STARTING THE CACHE ================== The cache is started by running the daemon. The daemon opens the cache device, configures the cache and tells it to begin caching. At that point the cache binds to fscache and the cache becomes live. The daemon is run as follows: /sbin/cachefilesd [-d]* [-s] [-n] [-f <configfile>] The flags are: (*) -d Increase the debugging level. This can be specified multiple times and is cumulative with itself. (*) -s Send messages to stderr instead of syslog. (*) -n Don't daemonise and go into background. (*) -f <configfile> Use an alternative configuration file rather than the default one. =============== THINGS TO AVOID =============== Do not mount other things within the cache as this will cause problems. The kernel module contains its own very cut-down path walking facility that ignores mountpoints, but the daemon can't avoid them. Do not create, rename or unlink files and directories in the cache whilst the cache is active, as this may cause the state to become uncertain. Renaming files in the cache might make objects appear to be other objects (the filename is part of the lookup key). Do not change or remove the extended attributes attached to cache files by the cache as this will cause the cache state management to get confused. Do not create files or directories in the cache, lest the cache get confused or serve incorrect data. Do not chmod files in the cache. The module creates things with minimal permissions to prevent random users being able to access them directly. ============= CACHE CULLING ============= The cache may need culling occasionally to make space. This involves discarding objects from the cache that have been used less recently than anything else. Culling is based on the access time of data objects. Empty directories are culled if not in use. Cache culling is done on the basis of the percentage of blocks and the percentage of files available in the underlying filesystem. There are six "limits": (*) brun (*) frun If the amount of free space and the number of available files in the cache rises above both these limits, then culling is turned off. (*) bcull (*) fcull If the amount of available space or the number of available files in the cache falls below either of these limits, then culling is started. (*) bstop (*) fstop If the amount of available space or the number of available files in the cache falls below either of these limits, then no further allocation of disk space or files is permitted until culling has raised things above these limits again. These must be configured thusly: 0 <= bstop < bcull < brun < 100 0 <= fstop < fcull < frun < 100 Note that these are percentages of available space and available files, and do _not_ appear as 100 minus the percentage displayed by the "df" program. The userspace daemon scans the cache to build up a table of cullable objects. These are then culled in least recently used order. A new scan of the cache is started as soon as space is made in the table. Objects will be skipped if their atimes have changed or if the kernel module says it is still using them. =============== CACHE STRUCTURE =============== The CacheFiles module will create two directories in the directory it was given: (*) cache/ (*) graveyard/ The active cache objects all reside in the first directory. The CacheFiles kernel module moves any retired or culled objects that it can't simply unlink to the graveyard from which the daemon will actually delete them. The daemon uses dnotify to monitor the graveyard directory, and will delete anything that appears therein. The module represents index objects as directories with the filename "I..." or "J...". Note that the "cache/" directory is itself a special index. Data objects are represented as files if they have no children, or directories if they do. Their filenames all begin "D..." or "E...". If represented as a directory, data objects will have a file in the directory called "data" that actually holds the data. Special objects are similar to data objects, except their filenames begin "S..." or "T...". If an object has children, then it will be represented as a directory. Immediately in the representative directory are a collection of directories named for hash values of the child object keys with an '@' prepended. Into this directory, if possible, will be placed the representations of the child objects: INDEX INDEX INDEX DATA FILES ========= ========== ================================= ================ cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...DB1ry cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...N22ry cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...FP1ry If the key is so long that it exceeds NAME_MAX with the decorations added on to it, then it will be cut into pieces, the first few of which will be used to make a nest of directories, and the last one of which will be the objects inside the last directory. The names of the intermediate directories will have '+' prepended: J1223/@23/+xy...z/+kl...m/Epqr Note that keys are raw data, and not only may they exceed NAME_MAX in size, they may also contain things like '/' and NUL characters, and so they may not be suitable for turning directly into a filename. To handle this, CacheFiles will use a suitably printable filename directly and "base-64" encode ones that aren't directly suitable. The two versions of object filenames indicate the encoding: OBJECT TYPE PRINTABLE ENCODED =============== =============== =============== Index "I..." "J..." Data "D..." "E..." Special "S..." "T..." Intermediate directories are always "@" or "+" as appropriate. Each object in the cache has an extended attribute label that holds the object type ID (required to distinguish special objects) and the auxiliary data from the netfs. The latter is used to detect stale objects in the cache and update or retire them. Note that CacheFiles will erase from the cache any file it doesn't recognise or any file of an incorrect type (such as a FIFO file or a device file). ========================== SECURITY MODEL AND SELINUX ========================== CacheFiles is implemented to deal properly with the LSM security features of the Linux kernel and the SELinux facility. One of the problems that CacheFiles faces is that it is generally acting on behalf of a process, and running in that process's context, and that includes a security context that is not appropriate for accessing the cache - either because the files in the cache are inaccessible to that process, or because if the process creates a file in the cache, that file may be inaccessible to other processes. The way CacheFiles works is to temporarily change the security context (fsuid, fsgid and actor security label) that the process acts as - without changing the security context of the process when it the target of an operation performed by some other process (so signalling and suchlike still work correctly). When the CacheFiles module is asked to bind to its cache, it: (1) Finds the security label attached to the root cache directory and uses that as the security label with which it will create files. By default, this is: cachefiles_var_t (2) Finds the security label of the process which issued the bind request (presumed to be the cachefilesd daemon), which by default will be: cachefilesd_t and asks LSM to supply a security ID as which it should act given the daemon's label. By default, this will be: cachefiles_kernel_t SELinux transitions the daemon's security ID to the module's security ID based on a rule of this form in the policy. type_transition <daemon's-ID> kernel_t : process <module's-ID>; For instance: type_transition cachefilesd_t kernel_t : process cachefiles_kernel_t; The module's security ID gives it permission to create, move and remove files and directories in the cache, to find and access directories and files in the cache, to set and access extended attributes on cache objects, and to read and write files in the cache. The daemon's security ID gives it only a very restricted set of permissions: it may scan directories, stat files and erase files and directories. It may not read or write files in the cache, and so it is precluded from accessing the data cached therein; nor is it permitted to create new files in the cache. There are policy source files available in: http://people.redhat.com/~dhowells/fscache/cachefilesd-0.8.tar.bz2 and later versions. In that tarball, see the files: cachefilesd.te cachefilesd.fc cachefilesd.if They are built and installed directly by the RPM. If a non-RPM based system is being used, then copy the above files to their own directory and run: make -f /usr/share/selinux/devel/Makefile semodule -i cachefilesd.pp You will need checkpolicy and selinux-policy-devel installed prior to the build. By default, the cache is located in /var/fscache, but if it is desirable that it should be elsewhere, than either the above policy files must be altered, or an auxiliary policy must be installed to label the alternate location of the cache. For instructions on how to add an auxiliary policy to enable the cache to be located elsewhere when SELinux is in enforcing mode, please see: /usr/share/doc/cachefilesd-*/move-cache.txt When the cachefilesd rpm is installed; alternatively, the document can be found in the sources. ================== A NOTE ON SECURITY ================== CacheFiles makes use of the split security in the task_struct. It allocates its own task_security structure, and redirects current->act_as to point to it when it acts on behalf of another process, in that process's context. The reason it does this is that it calls vfs_mkdir() and suchlike rather than bypassing security and calling inode ops directly. Therefore the VFS and LSM may deny the CacheFiles access to the cache data because under some circumstances the caching code is running in the security context of whatever process issued the original syscall on the netfs. Furthermore, should CacheFiles create a file or directory, the security parameters with that object is created (UID, GID, security label) would be derived from that process that issued the system call, thus potentially preventing other processes from accessing the cache - including CacheFiles's cache management daemon (cachefilesd). What is required is to temporarily override the security of the process that issued the system call. We can't, however, just do an in-place change of the security data as that affects the process as an object, not just as a subject. This means it may lose signals or ptrace events for example, and affects what the process looks like in /proc. So CacheFiles makes use of a logical split in the security between the objective security (task->sec) and the subjective security (task->act_as). The objective security holds the intrinsic security properties of a process and is never overridden. This is what appears in /proc, and is what is used when a process is the target of an operation by some other process (SIGKILL for example). The subjective security holds the active security properties of a process, and may be overridden. This is not seen externally, and is used whan a process acts upon another object, for example SIGKILLing another process or opening a file. LSM hooks exist that allow SELinux (or Smack or whatever) to reject a request for CacheFiles to run in a context of a specific security label, or to create files and directories with another security label. This documentation is added by the patch to: Documentation/filesystems/caching/cachefiles.txt Signed-Off-By: David Howells <dhowells@redhat.com> Acked-by: Steve Dickson <steved@redhat.com> Acked-by: Trond Myklebust <Trond.Myklebust@netapp.com> Acked-by: Al Viro <viro@zeniv.linux.org.uk> Tested-by: Daire Byrne <Daire.Byrne@framestore.com>
Diffstat (limited to 'fs')
-rw-r--r--fs/Kconfig1
-rw-r--r--fs/Makefile1
-rw-r--r--fs/cachefiles/Kconfig39
-rw-r--r--fs/cachefiles/Makefile18
-rw-r--r--fs/cachefiles/bind.c286
-rw-r--r--fs/cachefiles/daemon.c755
-rw-r--r--fs/cachefiles/interface.c449
-rw-r--r--fs/cachefiles/internal.h360
-rw-r--r--fs/cachefiles/key.c159
-rw-r--r--fs/cachefiles/main.c106
-rw-r--r--fs/cachefiles/namei.c771
-rw-r--r--fs/cachefiles/proc.c134
-rw-r--r--fs/cachefiles/rdwr.c879
-rw-r--r--fs/cachefiles/security.c116
-rw-r--r--fs/cachefiles/xattr.c291
15 files changed, 4365 insertions, 0 deletions
diff --git a/fs/Kconfig b/fs/Kconfig
index 3942df6ad04f..c0022b1d5877 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -69,6 +69,7 @@ config GENERIC_ACL
69menu "Caches" 69menu "Caches"
70 70
71source "fs/fscache/Kconfig" 71source "fs/fscache/Kconfig"
72source "fs/cachefiles/Kconfig"
72 73
73endmenu 74endmenu
74 75
diff --git a/fs/Makefile b/fs/Makefile
index ff5a10d898c6..055d5237b109 100644
--- a/fs/Makefile
+++ b/fs/Makefile
@@ -117,6 +117,7 @@ obj-$(CONFIG_AFS_FS) += afs/
117obj-$(CONFIG_BEFS_FS) += befs/ 117obj-$(CONFIG_BEFS_FS) += befs/
118obj-$(CONFIG_HOSTFS) += hostfs/ 118obj-$(CONFIG_HOSTFS) += hostfs/
119obj-$(CONFIG_HPPFS) += hppfs/ 119obj-$(CONFIG_HPPFS) += hppfs/
120obj-$(CONFIG_CACHEFILES) += cachefiles/
120obj-$(CONFIG_DEBUG_FS) += debugfs/ 121obj-$(CONFIG_DEBUG_FS) += debugfs/
121obj-$(CONFIG_OCFS2_FS) += ocfs2/ 122obj-$(CONFIG_OCFS2_FS) += ocfs2/
122obj-$(CONFIG_BTRFS_FS) += btrfs/ 123obj-$(CONFIG_BTRFS_FS) += btrfs/
diff --git a/fs/cachefiles/Kconfig b/fs/cachefiles/Kconfig
new file mode 100644
index 000000000000..80e9c6167f0b
--- /dev/null
+++ b/fs/cachefiles/Kconfig
@@ -0,0 +1,39 @@
1
2config CACHEFILES
3 tristate "Filesystem caching on files"
4 depends on FSCACHE && BLOCK
5 help
6 This permits use of a mounted filesystem as a cache for other
7 filesystems - primarily networking filesystems - thus allowing fast
8 local disk to enhance the speed of slower devices.
9
10 See Documentation/filesystems/caching/cachefiles.txt for more
11 information.
12
13config CACHEFILES_DEBUG
14 bool "Debug CacheFiles"
15 depends on CACHEFILES
16 help
17 This permits debugging to be dynamically enabled in the filesystem
18 caching on files module. If this is set, the debugging output may be
19 enabled by setting bits in /sys/modules/cachefiles/parameter/debug or
20 by including a debugging specifier in /etc/cachefilesd.conf.
21
22config CACHEFILES_HISTOGRAM
23 bool "Gather latency information on CacheFiles"
24 depends on CACHEFILES && PROC_FS
25 help
26
27 This option causes latency information to be gathered on CacheFiles
28 operation and exported through file:
29
30 /proc/fs/cachefiles/histogram
31
32 The generation of this histogram adds a certain amount of overhead to
33 execution as there are a number of points at which data is gathered,
34 and on a multi-CPU system these may be on cachelines that keep
35 bouncing between CPUs. On the other hand, the histogram may be
36 useful for debugging purposes. Saying 'N' here is recommended.
37
38 See Documentation/filesystems/caching/cachefiles.txt for more
39 information.
diff --git a/fs/cachefiles/Makefile b/fs/cachefiles/Makefile
new file mode 100644
index 000000000000..32cbab0ffce3
--- /dev/null
+++ b/fs/cachefiles/Makefile
@@ -0,0 +1,18 @@
1#
2# Makefile for caching in a mounted filesystem
3#
4
5cachefiles-y := \
6 bind.o \
7 daemon.o \
8 interface.o \
9 key.o \
10 main.o \
11 namei.o \
12 rdwr.o \
13 security.o \
14 xattr.o
15
16cachefiles-$(CONFIG_CACHEFILES_HISTOGRAM) += proc.o
17
18obj-$(CONFIG_CACHEFILES) := cachefiles.o
diff --git a/fs/cachefiles/bind.c b/fs/cachefiles/bind.c
new file mode 100644
index 000000000000..3797e0077b35
--- /dev/null
+++ b/fs/cachefiles/bind.c
@@ -0,0 +1,286 @@
1/* Bind and unbind a cache from the filesystem backing it
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/module.h>
13#include <linux/init.h>
14#include <linux/sched.h>
15#include <linux/completion.h>
16#include <linux/slab.h>
17#include <linux/fs.h>
18#include <linux/file.h>
19#include <linux/namei.h>
20#include <linux/mount.h>
21#include <linux/statfs.h>
22#include <linux/ctype.h>
23#include "internal.h"
24
25static int cachefiles_daemon_add_cache(struct cachefiles_cache *caches);
26
27/*
28 * bind a directory as a cache
29 */
30int cachefiles_daemon_bind(struct cachefiles_cache *cache, char *args)
31{
32 _enter("{%u,%u,%u,%u,%u,%u},%s",
33 cache->frun_percent,
34 cache->fcull_percent,
35 cache->fstop_percent,
36 cache->brun_percent,
37 cache->bcull_percent,
38 cache->bstop_percent,
39 args);
40
41 /* start by checking things over */
42 ASSERT(cache->fstop_percent >= 0 &&
43 cache->fstop_percent < cache->fcull_percent &&
44 cache->fcull_percent < cache->frun_percent &&
45 cache->frun_percent < 100);
46
47 ASSERT(cache->bstop_percent >= 0 &&
48 cache->bstop_percent < cache->bcull_percent &&
49 cache->bcull_percent < cache->brun_percent &&
50 cache->brun_percent < 100);
51
52 if (*args) {
53 kerror("'bind' command doesn't take an argument");
54 return -EINVAL;
55 }
56
57 if (!cache->rootdirname) {
58 kerror("No cache directory specified");
59 return -EINVAL;
60 }
61
62 /* don't permit already bound caches to be re-bound */
63 if (test_bit(CACHEFILES_READY, &cache->flags)) {
64 kerror("Cache already bound");
65 return -EBUSY;
66 }
67
68 /* make sure we have copies of the tag and dirname strings */
69 if (!cache->tag) {
70 /* the tag string is released by the fops->release()
71 * function, so we don't release it on error here */
72 cache->tag = kstrdup("CacheFiles", GFP_KERNEL);
73 if (!cache->tag)
74 return -ENOMEM;
75 }
76
77 /* add the cache */
78 return cachefiles_daemon_add_cache(cache);
79}
80
81/*
82 * add a cache
83 */
84static int cachefiles_daemon_add_cache(struct cachefiles_cache *cache)
85{
86 struct cachefiles_object *fsdef;
87 struct nameidata nd;
88 struct kstatfs stats;
89 struct dentry *graveyard, *cachedir, *root;
90 const struct cred *saved_cred;
91 int ret;
92
93 _enter("");
94
95 /* we want to work under the module's security ID */
96 ret = cachefiles_get_security_ID(cache);
97 if (ret < 0)
98 return ret;
99
100 cachefiles_begin_secure(cache, &saved_cred);
101
102 /* allocate the root index object */
103 ret = -ENOMEM;
104
105 fsdef = kmem_cache_alloc(cachefiles_object_jar, GFP_KERNEL);
106 if (!fsdef)
107 goto error_root_object;
108
109 ASSERTCMP(fsdef->backer, ==, NULL);
110
111 atomic_set(&fsdef->usage, 1);
112 fsdef->type = FSCACHE_COOKIE_TYPE_INDEX;
113
114 _debug("- fsdef %p", fsdef);
115
116 /* look up the directory at the root of the cache */
117 memset(&nd, 0, sizeof(nd));
118
119 ret = path_lookup(cache->rootdirname, LOOKUP_DIRECTORY, &nd);
120 if (ret < 0)
121 goto error_open_root;
122
123 cache->mnt = mntget(nd.path.mnt);
124 root = dget(nd.path.dentry);
125 path_put(&nd.path);
126
127 /* check parameters */
128 ret = -EOPNOTSUPP;
129 if (!root->d_inode ||
130 !root->d_inode->i_op ||
131 !root->d_inode->i_op->lookup ||
132 !root->d_inode->i_op->mkdir ||
133 !root->d_inode->i_op->setxattr ||
134 !root->d_inode->i_op->getxattr ||
135 !root->d_sb ||
136 !root->d_sb->s_op ||
137 !root->d_sb->s_op->statfs ||
138 !root->d_sb->s_op->sync_fs)
139 goto error_unsupported;
140
141 ret = -EROFS;
142 if (root->d_sb->s_flags & MS_RDONLY)
143 goto error_unsupported;
144
145 /* determine the security of the on-disk cache as this governs
146 * security ID of files we create */
147 ret = cachefiles_determine_cache_security(cache, root, &saved_cred);
148 if (ret < 0)
149 goto error_unsupported;
150
151 /* get the cache size and blocksize */
152 ret = vfs_statfs(root, &stats);
153 if (ret < 0)
154 goto error_unsupported;
155
156 ret = -ERANGE;
157 if (stats.f_bsize <= 0)
158 goto error_unsupported;
159
160 ret = -EOPNOTSUPP;
161 if (stats.f_bsize > PAGE_SIZE)
162 goto error_unsupported;
163
164 cache->bsize = stats.f_bsize;
165 cache->bshift = 0;
166 if (stats.f_bsize < PAGE_SIZE)
167 cache->bshift = PAGE_SHIFT - ilog2(stats.f_bsize);
168
169 _debug("blksize %u (shift %u)",
170 cache->bsize, cache->bshift);
171
172 _debug("size %llu, avail %llu",
173 (unsigned long long) stats.f_blocks,
174 (unsigned long long) stats.f_bavail);
175
176 /* set up caching limits */
177 do_div(stats.f_files, 100);
178 cache->fstop = stats.f_files * cache->fstop_percent;
179 cache->fcull = stats.f_files * cache->fcull_percent;
180 cache->frun = stats.f_files * cache->frun_percent;
181
182 _debug("limits {%llu,%llu,%llu} files",
183 (unsigned long long) cache->frun,
184 (unsigned long long) cache->fcull,
185 (unsigned long long) cache->fstop);
186
187 stats.f_blocks >>= cache->bshift;
188 do_div(stats.f_blocks, 100);
189 cache->bstop = stats.f_blocks * cache->bstop_percent;
190 cache->bcull = stats.f_blocks * cache->bcull_percent;
191 cache->brun = stats.f_blocks * cache->brun_percent;
192
193 _debug("limits {%llu,%llu,%llu} blocks",
194 (unsigned long long) cache->brun,
195 (unsigned long long) cache->bcull,
196 (unsigned long long) cache->bstop);
197
198 /* get the cache directory and check its type */
199 cachedir = cachefiles_get_directory(cache, root, "cache");
200 if (IS_ERR(cachedir)) {
201 ret = PTR_ERR(cachedir);
202 goto error_unsupported;
203 }
204
205 fsdef->dentry = cachedir;
206 fsdef->fscache.cookie = NULL;
207
208 ret = cachefiles_check_object_type(fsdef);
209 if (ret < 0)
210 goto error_unsupported;
211
212 /* get the graveyard directory */
213 graveyard = cachefiles_get_directory(cache, root, "graveyard");
214 if (IS_ERR(graveyard)) {
215 ret = PTR_ERR(graveyard);
216 goto error_unsupported;
217 }
218
219 cache->graveyard = graveyard;
220
221 /* publish the cache */
222 fscache_init_cache(&cache->cache,
223 &cachefiles_cache_ops,
224 "%s",
225 fsdef->dentry->d_sb->s_id);
226
227 fscache_object_init(&fsdef->fscache, NULL, &cache->cache);
228
229 ret = fscache_add_cache(&cache->cache, &fsdef->fscache, cache->tag);
230 if (ret < 0)
231 goto error_add_cache;
232
233 /* done */
234 set_bit(CACHEFILES_READY, &cache->flags);
235 dput(root);
236
237 printk(KERN_INFO "CacheFiles:"
238 " File cache on %s registered\n",
239 cache->cache.identifier);
240
241 /* check how much space the cache has */
242 cachefiles_has_space(cache, 0, 0);
243 cachefiles_end_secure(cache, saved_cred);
244 return 0;
245
246error_add_cache:
247 dput(cache->graveyard);
248 cache->graveyard = NULL;
249error_unsupported:
250 mntput(cache->mnt);
251 cache->mnt = NULL;
252 dput(fsdef->dentry);
253 fsdef->dentry = NULL;
254 dput(root);
255error_open_root:
256 kmem_cache_free(cachefiles_object_jar, fsdef);
257error_root_object:
258 cachefiles_end_secure(cache, saved_cred);
259 kerror("Failed to register: %d", ret);
260 return ret;
261}
262
263/*
264 * unbind a cache on fd release
265 */
266void cachefiles_daemon_unbind(struct cachefiles_cache *cache)
267{
268 _enter("");
269
270 if (test_bit(CACHEFILES_READY, &cache->flags)) {
271 printk(KERN_INFO "CacheFiles:"
272 " File cache on %s unregistering\n",
273 cache->cache.identifier);
274
275 fscache_withdraw_cache(&cache->cache);
276 }
277
278 dput(cache->graveyard);
279 mntput(cache->mnt);
280
281 kfree(cache->rootdirname);
282 kfree(cache->secctx);
283 kfree(cache->tag);
284
285 _leave("");
286}
diff --git a/fs/cachefiles/daemon.c b/fs/cachefiles/daemon.c
new file mode 100644
index 000000000000..4618516dd994
--- /dev/null
+++ b/fs/cachefiles/daemon.c
@@ -0,0 +1,755 @@
1/* Daemon interface
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/module.h>
13#include <linux/init.h>
14#include <linux/sched.h>
15#include <linux/completion.h>
16#include <linux/slab.h>
17#include <linux/fs.h>
18#include <linux/file.h>
19#include <linux/namei.h>
20#include <linux/poll.h>
21#include <linux/mount.h>
22#include <linux/statfs.h>
23#include <linux/ctype.h>
24#include <linux/fs_struct.h>
25#include "internal.h"
26
27static int cachefiles_daemon_open(struct inode *, struct file *);
28static int cachefiles_daemon_release(struct inode *, struct file *);
29static ssize_t cachefiles_daemon_read(struct file *, char __user *, size_t,
30 loff_t *);
31static ssize_t cachefiles_daemon_write(struct file *, const char __user *,
32 size_t, loff_t *);
33static unsigned int cachefiles_daemon_poll(struct file *,
34 struct poll_table_struct *);
35static int cachefiles_daemon_frun(struct cachefiles_cache *, char *);
36static int cachefiles_daemon_fcull(struct cachefiles_cache *, char *);
37static int cachefiles_daemon_fstop(struct cachefiles_cache *, char *);
38static int cachefiles_daemon_brun(struct cachefiles_cache *, char *);
39static int cachefiles_daemon_bcull(struct cachefiles_cache *, char *);
40static int cachefiles_daemon_bstop(struct cachefiles_cache *, char *);
41static int cachefiles_daemon_cull(struct cachefiles_cache *, char *);
42static int cachefiles_daemon_debug(struct cachefiles_cache *, char *);
43static int cachefiles_daemon_dir(struct cachefiles_cache *, char *);
44static int cachefiles_daemon_inuse(struct cachefiles_cache *, char *);
45static int cachefiles_daemon_secctx(struct cachefiles_cache *, char *);
46static int cachefiles_daemon_tag(struct cachefiles_cache *, char *);
47
48static unsigned long cachefiles_open;
49
50const struct file_operations cachefiles_daemon_fops = {
51 .owner = THIS_MODULE,
52 .open = cachefiles_daemon_open,
53 .release = cachefiles_daemon_release,
54 .read = cachefiles_daemon_read,
55 .write = cachefiles_daemon_write,
56 .poll = cachefiles_daemon_poll,
57};
58
59struct cachefiles_daemon_cmd {
60 char name[8];
61 int (*handler)(struct cachefiles_cache *cache, char *args);
62};
63
64static const struct cachefiles_daemon_cmd cachefiles_daemon_cmds[] = {
65 { "bind", cachefiles_daemon_bind },
66 { "brun", cachefiles_daemon_brun },
67 { "bcull", cachefiles_daemon_bcull },
68 { "bstop", cachefiles_daemon_bstop },
69 { "cull", cachefiles_daemon_cull },
70 { "debug", cachefiles_daemon_debug },
71 { "dir", cachefiles_daemon_dir },
72 { "frun", cachefiles_daemon_frun },
73 { "fcull", cachefiles_daemon_fcull },
74 { "fstop", cachefiles_daemon_fstop },
75 { "inuse", cachefiles_daemon_inuse },
76 { "secctx", cachefiles_daemon_secctx },
77 { "tag", cachefiles_daemon_tag },
78 { "", NULL }
79};
80
81
82/*
83 * do various checks
84 */
85static int cachefiles_daemon_open(struct inode *inode, struct file *file)
86{
87 struct cachefiles_cache *cache;
88
89 _enter("");
90
91 /* only the superuser may do this */
92 if (!capable(CAP_SYS_ADMIN))
93 return -EPERM;
94
95 /* the cachefiles device may only be open once at a time */
96 if (xchg(&cachefiles_open, 1) == 1)
97 return -EBUSY;
98
99 /* allocate a cache record */
100 cache = kzalloc(sizeof(struct cachefiles_cache), GFP_KERNEL);
101 if (!cache) {
102 cachefiles_open = 0;
103 return -ENOMEM;
104 }
105
106 mutex_init(&cache->daemon_mutex);
107 cache->active_nodes = RB_ROOT;
108 rwlock_init(&cache->active_lock);
109 init_waitqueue_head(&cache->daemon_pollwq);
110
111 /* set default caching limits
112 * - limit at 1% free space and/or free files
113 * - cull below 5% free space and/or free files
114 * - cease culling above 7% free space and/or free files
115 */
116 cache->frun_percent = 7;
117 cache->fcull_percent = 5;
118 cache->fstop_percent = 1;
119 cache->brun_percent = 7;
120 cache->bcull_percent = 5;
121 cache->bstop_percent = 1;
122
123 file->private_data = cache;
124 cache->cachefilesd = file;
125 return 0;
126}
127
128/*
129 * release a cache
130 */
131static int cachefiles_daemon_release(struct inode *inode, struct file *file)
132{
133 struct cachefiles_cache *cache = file->private_data;
134
135 _enter("");
136
137 ASSERT(cache);
138
139 set_bit(CACHEFILES_DEAD, &cache->flags);
140
141 cachefiles_daemon_unbind(cache);
142
143 ASSERT(!cache->active_nodes.rb_node);
144
145 /* clean up the control file interface */
146 cache->cachefilesd = NULL;
147 file->private_data = NULL;
148 cachefiles_open = 0;
149
150 kfree(cache);
151
152 _leave("");
153 return 0;
154}
155
156/*
157 * read the cache state
158 */
159static ssize_t cachefiles_daemon_read(struct file *file, char __user *_buffer,
160 size_t buflen, loff_t *pos)
161{
162 struct cachefiles_cache *cache = file->private_data;
163 char buffer[256];
164 int n;
165
166 //_enter(",,%zu,", buflen);
167
168 if (!test_bit(CACHEFILES_READY, &cache->flags))
169 return 0;
170
171 /* check how much space the cache has */
172 cachefiles_has_space(cache, 0, 0);
173
174 /* summarise */
175 clear_bit(CACHEFILES_STATE_CHANGED, &cache->flags);
176
177 n = snprintf(buffer, sizeof(buffer),
178 "cull=%c"
179 " frun=%llx"
180 " fcull=%llx"
181 " fstop=%llx"
182 " brun=%llx"
183 " bcull=%llx"
184 " bstop=%llx",
185 test_bit(CACHEFILES_CULLING, &cache->flags) ? '1' : '0',
186 (unsigned long long) cache->frun,
187 (unsigned long long) cache->fcull,
188 (unsigned long long) cache->fstop,
189 (unsigned long long) cache->brun,
190 (unsigned long long) cache->bcull,
191 (unsigned long long) cache->bstop
192 );
193
194 if (n > buflen)
195 return -EMSGSIZE;
196
197 if (copy_to_user(_buffer, buffer, n) != 0)
198 return -EFAULT;
199
200 return n;
201}
202
203/*
204 * command the cache
205 */
206static ssize_t cachefiles_daemon_write(struct file *file,
207 const char __user *_data,
208 size_t datalen,
209 loff_t *pos)
210{
211 const struct cachefiles_daemon_cmd *cmd;
212 struct cachefiles_cache *cache = file->private_data;
213 ssize_t ret;
214 char *data, *args, *cp;
215
216 //_enter(",,%zu,", datalen);
217
218 ASSERT(cache);
219
220 if (test_bit(CACHEFILES_DEAD, &cache->flags))
221 return -EIO;
222
223 if (datalen < 0 || datalen > PAGE_SIZE - 1)
224 return -EOPNOTSUPP;
225
226 /* drag the command string into the kernel so we can parse it */
227 data = kmalloc(datalen + 1, GFP_KERNEL);
228 if (!data)
229 return -ENOMEM;
230
231 ret = -EFAULT;
232 if (copy_from_user(data, _data, datalen) != 0)
233 goto error;
234
235 data[datalen] = '\0';
236
237 ret = -EINVAL;
238 if (memchr(data, '\0', datalen))
239 goto error;
240
241 /* strip any newline */
242 cp = memchr(data, '\n', datalen);
243 if (cp) {
244 if (cp == data)
245 goto error;
246
247 *cp = '\0';
248 }
249
250 /* parse the command */
251 ret = -EOPNOTSUPP;
252
253 for (args = data; *args; args++)
254 if (isspace(*args))
255 break;
256 if (*args) {
257 if (args == data)
258 goto error;
259 *args = '\0';
260 for (args++; isspace(*args); args++)
261 continue;
262 }
263
264 /* run the appropriate command handler */
265 for (cmd = cachefiles_daemon_cmds; cmd->name[0]; cmd++)
266 if (strcmp(cmd->name, data) == 0)
267 goto found_command;
268
269error:
270 kfree(data);
271 //_leave(" = %zd", ret);
272 return ret;
273
274found_command:
275 mutex_lock(&cache->daemon_mutex);
276
277 ret = -EIO;
278 if (!test_bit(CACHEFILES_DEAD, &cache->flags))
279 ret = cmd->handler(cache, args);
280
281 mutex_unlock(&cache->daemon_mutex);
282
283 if (ret == 0)
284 ret = datalen;
285 goto error;
286}
287
288/*
289 * poll for culling state
290 * - use POLLOUT to indicate culling state
291 */
292static unsigned int cachefiles_daemon_poll(struct file *file,
293 struct poll_table_struct *poll)
294{
295 struct cachefiles_cache *cache = file->private_data;
296 unsigned int mask;
297
298 poll_wait(file, &cache->daemon_pollwq, poll);
299 mask = 0;
300
301 if (test_bit(CACHEFILES_STATE_CHANGED, &cache->flags))
302 mask |= POLLIN;
303
304 if (test_bit(CACHEFILES_CULLING, &cache->flags))
305 mask |= POLLOUT;
306
307 return mask;
308}
309
310/*
311 * give a range error for cache space constraints
312 * - can be tail-called
313 */
314static int cachefiles_daemon_range_error(struct cachefiles_cache *cache,
315 char *args)
316{
317 kerror("Free space limits must be in range"
318 " 0%%<=stop<cull<run<100%%");
319
320 return -EINVAL;
321}
322
323/*
324 * set the percentage of files at which to stop culling
325 * - command: "frun <N>%"
326 */
327static int cachefiles_daemon_frun(struct cachefiles_cache *cache, char *args)
328{
329 unsigned long frun;
330
331 _enter(",%s", args);
332
333 if (!*args)
334 return -EINVAL;
335
336 frun = simple_strtoul(args, &args, 10);
337 if (args[0] != '%' || args[1] != '\0')
338 return -EINVAL;
339
340 if (frun <= cache->fcull_percent || frun >= 100)
341 return cachefiles_daemon_range_error(cache, args);
342
343 cache->frun_percent = frun;
344 return 0;
345}
346
347/*
348 * set the percentage of files at which to start culling
349 * - command: "fcull <N>%"
350 */
351static int cachefiles_daemon_fcull(struct cachefiles_cache *cache, char *args)
352{
353 unsigned long fcull;
354
355 _enter(",%s", args);
356
357 if (!*args)
358 return -EINVAL;
359
360 fcull = simple_strtoul(args, &args, 10);
361 if (args[0] != '%' || args[1] != '\0')
362 return -EINVAL;
363
364 if (fcull <= cache->fstop_percent || fcull >= cache->frun_percent)
365 return cachefiles_daemon_range_error(cache, args);
366
367 cache->fcull_percent = fcull;
368 return 0;
369}
370
371/*
372 * set the percentage of files at which to stop allocating
373 * - command: "fstop <N>%"
374 */
375static int cachefiles_daemon_fstop(struct cachefiles_cache *cache, char *args)
376{
377 unsigned long fstop;
378
379 _enter(",%s", args);
380
381 if (!*args)
382 return -EINVAL;
383
384 fstop = simple_strtoul(args, &args, 10);
385 if (args[0] != '%' || args[1] != '\0')
386 return -EINVAL;
387
388 if (fstop < 0 || fstop >= cache->fcull_percent)
389 return cachefiles_daemon_range_error(cache, args);
390
391 cache->fstop_percent = fstop;
392 return 0;
393}
394
395/*
396 * set the percentage of blocks at which to stop culling
397 * - command: "brun <N>%"
398 */
399static int cachefiles_daemon_brun(struct cachefiles_cache *cache, char *args)
400{
401 unsigned long brun;
402
403 _enter(",%s", args);
404
405 if (!*args)
406 return -EINVAL;
407
408 brun = simple_strtoul(args, &args, 10);
409 if (args[0] != '%' || args[1] != '\0')
410 return -EINVAL;
411
412 if (brun <= cache->bcull_percent || brun >= 100)
413 return cachefiles_daemon_range_error(cache, args);
414
415 cache->brun_percent = brun;
416 return 0;
417}
418
419/*
420 * set the percentage of blocks at which to start culling
421 * - command: "bcull <N>%"
422 */
423static int cachefiles_daemon_bcull(struct cachefiles_cache *cache, char *args)
424{
425 unsigned long bcull;
426
427 _enter(",%s", args);
428
429 if (!*args)
430 return -EINVAL;
431
432 bcull = simple_strtoul(args, &args, 10);
433 if (args[0] != '%' || args[1] != '\0')
434 return -EINVAL;
435
436 if (bcull <= cache->bstop_percent || bcull >= cache->brun_percent)
437 return cachefiles_daemon_range_error(cache, args);
438
439 cache->bcull_percent = bcull;
440 return 0;
441}
442
443/*
444 * set the percentage of blocks at which to stop allocating
445 * - command: "bstop <N>%"
446 */
447static int cachefiles_daemon_bstop(struct cachefiles_cache *cache, char *args)
448{
449 unsigned long bstop;
450
451 _enter(",%s", args);
452
453 if (!*args)
454 return -EINVAL;
455
456 bstop = simple_strtoul(args, &args, 10);
457 if (args[0] != '%' || args[1] != '\0')
458 return -EINVAL;
459
460 if (bstop < 0 || bstop >= cache->bcull_percent)
461 return cachefiles_daemon_range_error(cache, args);
462
463 cache->bstop_percent = bstop;
464 return 0;
465}
466
467/*
468 * set the cache directory
469 * - command: "dir <name>"
470 */
471static int cachefiles_daemon_dir(struct cachefiles_cache *cache, char *args)
472{
473 char *dir;
474
475 _enter(",%s", args);
476
477 if (!*args) {
478 kerror("Empty directory specified");
479 return -EINVAL;
480 }
481
482 if (cache->rootdirname) {
483 kerror("Second cache directory specified");
484 return -EEXIST;
485 }
486
487 dir = kstrdup(args, GFP_KERNEL);
488 if (!dir)
489 return -ENOMEM;
490
491 cache->rootdirname = dir;
492 return 0;
493}
494
495/*
496 * set the cache security context
497 * - command: "secctx <ctx>"
498 */
499static int cachefiles_daemon_secctx(struct cachefiles_cache *cache, char *args)
500{
501 char *secctx;
502
503 _enter(",%s", args);
504
505 if (!*args) {
506 kerror("Empty security context specified");
507 return -EINVAL;
508 }
509
510 if (cache->secctx) {
511 kerror("Second security context specified");
512 return -EINVAL;
513 }
514
515 secctx = kstrdup(args, GFP_KERNEL);
516 if (!secctx)
517 return -ENOMEM;
518
519 cache->secctx = secctx;
520 return 0;
521}
522
523/*
524 * set the cache tag
525 * - command: "tag <name>"
526 */
527static int cachefiles_daemon_tag(struct cachefiles_cache *cache, char *args)
528{
529 char *tag;
530
531 _enter(",%s", args);
532
533 if (!*args) {
534 kerror("Empty tag specified");
535 return -EINVAL;
536 }
537
538 if (cache->tag)
539 return -EEXIST;
540
541 tag = kstrdup(args, GFP_KERNEL);
542 if (!tag)
543 return -ENOMEM;
544
545 cache->tag = tag;
546 return 0;
547}
548
549/*
550 * request a node in the cache be culled from the current working directory
551 * - command: "cull <name>"
552 */
553static int cachefiles_daemon_cull(struct cachefiles_cache *cache, char *args)
554{
555 struct fs_struct *fs;
556 struct dentry *dir;
557 const struct cred *saved_cred;
558 int ret;
559
560 _enter(",%s", args);
561
562 if (strchr(args, '/'))
563 goto inval;
564
565 if (!test_bit(CACHEFILES_READY, &cache->flags)) {
566 kerror("cull applied to unready cache");
567 return -EIO;
568 }
569
570 if (test_bit(CACHEFILES_DEAD, &cache->flags)) {
571 kerror("cull applied to dead cache");
572 return -EIO;
573 }
574
575 /* extract the directory dentry from the cwd */
576 fs = current->fs;
577 read_lock(&fs->lock);
578 dir = dget(fs->pwd.dentry);
579 read_unlock(&fs->lock);
580
581 if (!S_ISDIR(dir->d_inode->i_mode))
582 goto notdir;
583
584 cachefiles_begin_secure(cache, &saved_cred);
585 ret = cachefiles_cull(cache, dir, args);
586 cachefiles_end_secure(cache, saved_cred);
587
588 dput(dir);
589 _leave(" = %d", ret);
590 return ret;
591
592notdir:
593 dput(dir);
594 kerror("cull command requires dirfd to be a directory");
595 return -ENOTDIR;
596
597inval:
598 kerror("cull command requires dirfd and filename");
599 return -EINVAL;
600}
601
602/*
603 * set debugging mode
604 * - command: "debug <mask>"
605 */
606static int cachefiles_daemon_debug(struct cachefiles_cache *cache, char *args)
607{
608 unsigned long mask;
609
610 _enter(",%s", args);
611
612 mask = simple_strtoul(args, &args, 0);
613 if (args[0] != '\0')
614 goto inval;
615
616 cachefiles_debug = mask;
617 _leave(" = 0");
618 return 0;
619
620inval:
621 kerror("debug command requires mask");
622 return -EINVAL;
623}
624
625/*
626 * find out whether an object in the current working directory is in use or not
627 * - command: "inuse <name>"
628 */
629static int cachefiles_daemon_inuse(struct cachefiles_cache *cache, char *args)
630{
631 struct fs_struct *fs;
632 struct dentry *dir;
633 const struct cred *saved_cred;
634 int ret;
635
636 //_enter(",%s", args);
637
638 if (strchr(args, '/'))
639 goto inval;
640
641 if (!test_bit(CACHEFILES_READY, &cache->flags)) {
642 kerror("inuse applied to unready cache");
643 return -EIO;
644 }
645
646 if (test_bit(CACHEFILES_DEAD, &cache->flags)) {
647 kerror("inuse applied to dead cache");
648 return -EIO;
649 }
650
651 /* extract the directory dentry from the cwd */
652 fs = current->fs;
653 read_lock(&fs->lock);
654 dir = dget(fs->pwd.dentry);
655 read_unlock(&fs->lock);
656
657 if (!S_ISDIR(dir->d_inode->i_mode))
658 goto notdir;
659
660 cachefiles_begin_secure(cache, &saved_cred);
661 ret = cachefiles_check_in_use(cache, dir, args);
662 cachefiles_end_secure(cache, saved_cred);
663
664 dput(dir);
665 //_leave(" = %d", ret);
666 return ret;
667
668notdir:
669 dput(dir);
670 kerror("inuse command requires dirfd to be a directory");
671 return -ENOTDIR;
672
673inval:
674 kerror("inuse command requires dirfd and filename");
675 return -EINVAL;
676}
677
678/*
679 * see if we have space for a number of pages and/or a number of files in the
680 * cache
681 */
682int cachefiles_has_space(struct cachefiles_cache *cache,
683 unsigned fnr, unsigned bnr)
684{
685 struct kstatfs stats;
686 int ret;
687
688 //_enter("{%llu,%llu,%llu,%llu,%llu,%llu},%u,%u",
689 // (unsigned long long) cache->frun,
690 // (unsigned long long) cache->fcull,
691 // (unsigned long long) cache->fstop,
692 // (unsigned long long) cache->brun,
693 // (unsigned long long) cache->bcull,
694 // (unsigned long long) cache->bstop,
695 // fnr, bnr);
696
697 /* find out how many pages of blockdev are available */
698 memset(&stats, 0, sizeof(stats));
699
700 ret = vfs_statfs(cache->mnt->mnt_root, &stats);
701 if (ret < 0) {
702 if (ret == -EIO)
703 cachefiles_io_error(cache, "statfs failed");
704 _leave(" = %d", ret);
705 return ret;
706 }
707
708 stats.f_bavail >>= cache->bshift;
709
710 //_debug("avail %llu,%llu",
711 // (unsigned long long) stats.f_ffree,
712 // (unsigned long long) stats.f_bavail);
713
714 /* see if there is sufficient space */
715 if (stats.f_ffree > fnr)
716 stats.f_ffree -= fnr;
717 else
718 stats.f_ffree = 0;
719
720 if (stats.f_bavail > bnr)
721 stats.f_bavail -= bnr;
722 else
723 stats.f_bavail = 0;
724
725 ret = -ENOBUFS;
726 if (stats.f_ffree < cache->fstop ||
727 stats.f_bavail < cache->bstop)
728 goto begin_cull;
729
730 ret = 0;
731 if (stats.f_ffree < cache->fcull ||
732 stats.f_bavail < cache->bcull)
733 goto begin_cull;
734
735 if (test_bit(CACHEFILES_CULLING, &cache->flags) &&
736 stats.f_ffree >= cache->frun &&
737 stats.f_bavail >= cache->brun &&
738 test_and_clear_bit(CACHEFILES_CULLING, &cache->flags)
739 ) {
740 _debug("cease culling");
741 cachefiles_state_changed(cache);
742 }
743
744 //_leave(" = 0");
745 return 0;
746
747begin_cull:
748 if (!test_and_set_bit(CACHEFILES_CULLING, &cache->flags)) {
749 _debug("### CULL CACHE ###");
750 cachefiles_state_changed(cache);
751 }
752
753 _leave(" = %d", ret);
754 return ret;
755}
diff --git a/fs/cachefiles/interface.c b/fs/cachefiles/interface.c
new file mode 100644
index 000000000000..1e962348d111
--- /dev/null
+++ b/fs/cachefiles/interface.c
@@ -0,0 +1,449 @@
1/* FS-Cache interface to CacheFiles
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/mount.h>
13#include <linux/buffer_head.h>
14#include "internal.h"
15
16#define list_to_page(head) (list_entry((head)->prev, struct page, lru))
17
18struct cachefiles_lookup_data {
19 struct cachefiles_xattr *auxdata; /* auxiliary data */
20 char *key; /* key path */
21};
22
23static int cachefiles_attr_changed(struct fscache_object *_object);
24
25/*
26 * allocate an object record for a cookie lookup and prepare the lookup data
27 */
28static struct fscache_object *cachefiles_alloc_object(
29 struct fscache_cache *_cache,
30 struct fscache_cookie *cookie)
31{
32 struct cachefiles_lookup_data *lookup_data;
33 struct cachefiles_object *object;
34 struct cachefiles_cache *cache;
35 struct cachefiles_xattr *auxdata;
36 unsigned keylen, auxlen;
37 void *buffer;
38 char *key;
39
40 cache = container_of(_cache, struct cachefiles_cache, cache);
41
42 _enter("{%s},%p,", cache->cache.identifier, cookie);
43
44 lookup_data = kmalloc(sizeof(*lookup_data), GFP_KERNEL);
45 if (!lookup_data)
46 goto nomem_lookup_data;
47
48 /* create a new object record and a temporary leaf image */
49 object = kmem_cache_alloc(cachefiles_object_jar, GFP_KERNEL);
50 if (!object)
51 goto nomem_object;
52
53 ASSERTCMP(object->backer, ==, NULL);
54
55 BUG_ON(test_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags));
56 atomic_set(&object->usage, 1);
57
58 fscache_object_init(&object->fscache, cookie, &cache->cache);
59
60 object->type = cookie->def->type;
61
62 /* get hold of the raw key
63 * - stick the length on the front and leave space on the back for the
64 * encoder
65 */
66 buffer = kmalloc((2 + 512) + 3, GFP_KERNEL);
67 if (!buffer)
68 goto nomem_buffer;
69
70 keylen = cookie->def->get_key(cookie->netfs_data, buffer + 2, 512);
71 ASSERTCMP(keylen, <, 512);
72
73 *(uint16_t *)buffer = keylen;
74 ((char *)buffer)[keylen + 2] = 0;
75 ((char *)buffer)[keylen + 3] = 0;
76 ((char *)buffer)[keylen + 4] = 0;
77
78 /* turn the raw key into something that can work with as a filename */
79 key = cachefiles_cook_key(buffer, keylen + 2, object->type);
80 if (!key)
81 goto nomem_key;
82
83 /* get hold of the auxiliary data and prepend the object type */
84 auxdata = buffer;
85 auxlen = 0;
86 if (cookie->def->get_aux) {
87 auxlen = cookie->def->get_aux(cookie->netfs_data,
88 auxdata->data, 511);
89 ASSERTCMP(auxlen, <, 511);
90 }
91
92 auxdata->len = auxlen + 1;
93 auxdata->type = cookie->def->type;
94
95 lookup_data->auxdata = auxdata;
96 lookup_data->key = key;
97 object->lookup_data = lookup_data;
98
99 _leave(" = %p [%p]", &object->fscache, lookup_data);
100 return &object->fscache;
101
102nomem_key:
103 kfree(buffer);
104nomem_buffer:
105 BUG_ON(test_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags));
106 kmem_cache_free(cachefiles_object_jar, object);
107 fscache_object_destroyed(&cache->cache);
108nomem_object:
109 kfree(lookup_data);
110nomem_lookup_data:
111 _leave(" = -ENOMEM");
112 return ERR_PTR(-ENOMEM);
113}
114
115/*
116 * attempt to look up the nominated node in this cache
117 */
118static void cachefiles_lookup_object(struct fscache_object *_object)
119{
120 struct cachefiles_lookup_data *lookup_data;
121 struct cachefiles_object *parent, *object;
122 struct cachefiles_cache *cache;
123 const struct cred *saved_cred;
124 int ret;
125
126 _enter("{OBJ%x}", _object->debug_id);
127
128 cache = container_of(_object->cache, struct cachefiles_cache, cache);
129 parent = container_of(_object->parent,
130 struct cachefiles_object, fscache);
131 object = container_of(_object, struct cachefiles_object, fscache);
132 lookup_data = object->lookup_data;
133
134 ASSERTCMP(lookup_data, !=, NULL);
135
136 /* look up the key, creating any missing bits */
137 cachefiles_begin_secure(cache, &saved_cred);
138 ret = cachefiles_walk_to_object(parent, object,
139 lookup_data->key,
140 lookup_data->auxdata);
141 cachefiles_end_secure(cache, saved_cred);
142
143 /* polish off by setting the attributes of non-index files */
144 if (ret == 0 &&
145 object->fscache.cookie->def->type != FSCACHE_COOKIE_TYPE_INDEX)
146 cachefiles_attr_changed(&object->fscache);
147
148 if (ret < 0) {
149 printk(KERN_WARNING "CacheFiles: Lookup failed error %d\n",
150 ret);
151 fscache_object_lookup_error(&object->fscache);
152 }
153
154 _leave(" [%d]", ret);
155}
156
157/*
158 * indication of lookup completion
159 */
160static void cachefiles_lookup_complete(struct fscache_object *_object)
161{
162 struct cachefiles_object *object;
163
164 object = container_of(_object, struct cachefiles_object, fscache);
165
166 _enter("{OBJ%x,%p}", object->fscache.debug_id, object->lookup_data);
167
168 if (object->lookup_data) {
169 kfree(object->lookup_data->key);
170 kfree(object->lookup_data->auxdata);
171 kfree(object->lookup_data);
172 object->lookup_data = NULL;
173 }
174}
175
176/*
177 * increment the usage count on an inode object (may fail if unmounting)
178 */
179static
180struct fscache_object *cachefiles_grab_object(struct fscache_object *_object)
181{
182 struct cachefiles_object *object =
183 container_of(_object, struct cachefiles_object, fscache);
184
185 _enter("{OBJ%x,%d}", _object->debug_id, atomic_read(&object->usage));
186
187#ifdef CACHEFILES_DEBUG_SLAB
188 ASSERT((atomic_read(&object->usage) & 0xffff0000) != 0x6b6b0000);
189#endif
190
191 atomic_inc(&object->usage);
192 return &object->fscache;
193}
194
195/*
196 * update the auxilliary data for an object object on disk
197 */
198static void cachefiles_update_object(struct fscache_object *_object)
199{
200 struct cachefiles_object *object;
201 struct cachefiles_xattr *auxdata;
202 struct cachefiles_cache *cache;
203 struct fscache_cookie *cookie;
204 const struct cred *saved_cred;
205 unsigned auxlen;
206
207 _enter("{OBJ%x}", _object->debug_id);
208
209 object = container_of(_object, struct cachefiles_object, fscache);
210 cache = container_of(object->fscache.cache, struct cachefiles_cache,
211 cache);
212 cookie = object->fscache.cookie;
213
214 if (!cookie->def->get_aux) {
215 _leave(" [no aux]");
216 return;
217 }
218
219 auxdata = kmalloc(2 + 512 + 3, GFP_KERNEL);
220 if (!auxdata) {
221 _leave(" [nomem]");
222 return;
223 }
224
225 auxlen = cookie->def->get_aux(cookie->netfs_data, auxdata->data, 511);
226 ASSERTCMP(auxlen, <, 511);
227
228 auxdata->len = auxlen + 1;
229 auxdata->type = cookie->def->type;
230
231 cachefiles_begin_secure(cache, &saved_cred);
232 cachefiles_update_object_xattr(object, auxdata);
233 cachefiles_end_secure(cache, saved_cred);
234 kfree(auxdata);
235 _leave("");
236}
237
238/*
239 * discard the resources pinned by an object and effect retirement if
240 * requested
241 */
242static void cachefiles_drop_object(struct fscache_object *_object)
243{
244 struct cachefiles_object *object;
245 struct cachefiles_cache *cache;
246 const struct cred *saved_cred;
247
248 ASSERT(_object);
249
250 object = container_of(_object, struct cachefiles_object, fscache);
251
252 _enter("{OBJ%x,%d}",
253 object->fscache.debug_id, atomic_read(&object->usage));
254
255 cache = container_of(object->fscache.cache,
256 struct cachefiles_cache, cache);
257
258#ifdef CACHEFILES_DEBUG_SLAB
259 ASSERT((atomic_read(&object->usage) & 0xffff0000) != 0x6b6b0000);
260#endif
261
262 /* delete retired objects */
263 if (object->fscache.state == FSCACHE_OBJECT_RECYCLING &&
264 _object != cache->cache.fsdef
265 ) {
266 _debug("- retire object OBJ%x", object->fscache.debug_id);
267 cachefiles_begin_secure(cache, &saved_cred);
268 cachefiles_delete_object(cache, object);
269 cachefiles_end_secure(cache, saved_cred);
270 }
271
272 /* close the filesystem stuff attached to the object */
273 if (object->backer != object->dentry)
274 dput(object->backer);
275 object->backer = NULL;
276
277 /* note that the object is now inactive */
278 if (test_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags)) {
279 write_lock(&cache->active_lock);
280 if (!test_and_clear_bit(CACHEFILES_OBJECT_ACTIVE,
281 &object->flags))
282 BUG();
283 rb_erase(&object->active_node, &cache->active_nodes);
284 wake_up_bit(&object->flags, CACHEFILES_OBJECT_ACTIVE);
285 write_unlock(&cache->active_lock);
286 }
287
288 dput(object->dentry);
289 object->dentry = NULL;
290
291 _leave("");
292}
293
294/*
295 * dispose of a reference to an object
296 */
297static void cachefiles_put_object(struct fscache_object *_object)
298{
299 struct cachefiles_object *object;
300 struct fscache_cache *cache;
301
302 ASSERT(_object);
303
304 object = container_of(_object, struct cachefiles_object, fscache);
305
306 _enter("{OBJ%x,%d}",
307 object->fscache.debug_id, atomic_read(&object->usage));
308
309#ifdef CACHEFILES_DEBUG_SLAB
310 ASSERT((atomic_read(&object->usage) & 0xffff0000) != 0x6b6b0000);
311#endif
312
313 ASSERTIFCMP(object->fscache.parent,
314 object->fscache.parent->n_children, >, 0);
315
316 if (atomic_dec_and_test(&object->usage)) {
317 _debug("- kill object OBJ%x", object->fscache.debug_id);
318
319 ASSERT(!test_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags));
320 ASSERTCMP(object->fscache.parent, ==, NULL);
321 ASSERTCMP(object->backer, ==, NULL);
322 ASSERTCMP(object->dentry, ==, NULL);
323 ASSERTCMP(object->fscache.n_ops, ==, 0);
324 ASSERTCMP(object->fscache.n_children, ==, 0);
325
326 if (object->lookup_data) {
327 kfree(object->lookup_data->key);
328 kfree(object->lookup_data->auxdata);
329 kfree(object->lookup_data);
330 object->lookup_data = NULL;
331 }
332
333 cache = object->fscache.cache;
334 kmem_cache_free(cachefiles_object_jar, object);
335 fscache_object_destroyed(cache);
336 }
337
338 _leave("");
339}
340
341/*
342 * sync a cache
343 */
344static void cachefiles_sync_cache(struct fscache_cache *_cache)
345{
346 struct cachefiles_cache *cache;
347 const struct cred *saved_cred;
348 int ret;
349
350 _enter("%p", _cache);
351
352 cache = container_of(_cache, struct cachefiles_cache, cache);
353
354 /* make sure all pages pinned by operations on behalf of the netfs are
355 * written to disc */
356 cachefiles_begin_secure(cache, &saved_cred);
357 ret = fsync_super(cache->mnt->mnt_sb);
358 cachefiles_end_secure(cache, saved_cred);
359
360 if (ret == -EIO)
361 cachefiles_io_error(cache,
362 "Attempt to sync backing fs superblock"
363 " returned error %d",
364 ret);
365}
366
367/*
368 * notification the attributes on an object have changed
369 * - called with reads/writes excluded by FS-Cache
370 */
371static int cachefiles_attr_changed(struct fscache_object *_object)
372{
373 struct cachefiles_object *object;
374 struct cachefiles_cache *cache;
375 const struct cred *saved_cred;
376 struct iattr newattrs;
377 uint64_t ni_size;
378 loff_t oi_size;
379 int ret;
380
381 _object->cookie->def->get_attr(_object->cookie->netfs_data, &ni_size);
382
383 _enter("{OBJ%x},[%llu]",
384 _object->debug_id, (unsigned long long) ni_size);
385
386 object = container_of(_object, struct cachefiles_object, fscache);
387 cache = container_of(object->fscache.cache,
388 struct cachefiles_cache, cache);
389
390 if (ni_size == object->i_size)
391 return 0;
392
393 if (!object->backer)
394 return -ENOBUFS;
395
396 ASSERT(S_ISREG(object->backer->d_inode->i_mode));
397
398 fscache_set_store_limit(&object->fscache, ni_size);
399
400 oi_size = i_size_read(object->backer->d_inode);
401 if (oi_size == ni_size)
402 return 0;
403
404 newattrs.ia_size = ni_size;
405 newattrs.ia_valid = ATTR_SIZE;
406
407 cachefiles_begin_secure(cache, &saved_cred);
408 mutex_lock(&object->backer->d_inode->i_mutex);
409 ret = notify_change(object->backer, &newattrs);
410 mutex_unlock(&object->backer->d_inode->i_mutex);
411 cachefiles_end_secure(cache, saved_cred);
412
413 if (ret == -EIO) {
414 fscache_set_store_limit(&object->fscache, 0);
415 cachefiles_io_error_obj(object, "Size set failed");
416 ret = -ENOBUFS;
417 }
418
419 _leave(" = %d", ret);
420 return ret;
421}
422
423/*
424 * dissociate a cache from all the pages it was backing
425 */
426static void cachefiles_dissociate_pages(struct fscache_cache *cache)
427{
428 _enter("");
429}
430
431const struct fscache_cache_ops cachefiles_cache_ops = {
432 .name = "cachefiles",
433 .alloc_object = cachefiles_alloc_object,
434 .lookup_object = cachefiles_lookup_object,
435 .lookup_complete = cachefiles_lookup_complete,
436 .grab_object = cachefiles_grab_object,
437 .update_object = cachefiles_update_object,
438 .drop_object = cachefiles_drop_object,
439 .put_object = cachefiles_put_object,
440 .sync_cache = cachefiles_sync_cache,
441 .attr_changed = cachefiles_attr_changed,
442 .read_or_alloc_page = cachefiles_read_or_alloc_page,
443 .read_or_alloc_pages = cachefiles_read_or_alloc_pages,
444 .allocate_page = cachefiles_allocate_page,
445 .allocate_pages = cachefiles_allocate_pages,
446 .write_page = cachefiles_write_page,
447 .uncache_page = cachefiles_uncache_page,
448 .dissociate_pages = cachefiles_dissociate_pages,
449};
diff --git a/fs/cachefiles/internal.h b/fs/cachefiles/internal.h
new file mode 100644
index 000000000000..19218e1463d6
--- /dev/null
+++ b/fs/cachefiles/internal.h
@@ -0,0 +1,360 @@
1/* General netfs cache on cache files internal defs
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/fscache-cache.h>
13#include <linux/timer.h>
14#include <linux/wait.h>
15#include <linux/workqueue.h>
16#include <linux/security.h>
17
18struct cachefiles_cache;
19struct cachefiles_object;
20
21extern unsigned cachefiles_debug;
22#define CACHEFILES_DEBUG_KENTER 1
23#define CACHEFILES_DEBUG_KLEAVE 2
24#define CACHEFILES_DEBUG_KDEBUG 4
25
26/*
27 * node records
28 */
29struct cachefiles_object {
30 struct fscache_object fscache; /* fscache handle */
31 struct cachefiles_lookup_data *lookup_data; /* cached lookup data */
32 struct dentry *dentry; /* the file/dir representing this object */
33 struct dentry *backer; /* backing file */
34 loff_t i_size; /* object size */
35 unsigned long flags;
36#define CACHEFILES_OBJECT_ACTIVE 0 /* T if marked active */
37 atomic_t usage; /* object usage count */
38 uint8_t type; /* object type */
39 uint8_t new; /* T if object new */
40 spinlock_t work_lock;
41 struct rb_node active_node; /* link in active tree (dentry is key) */
42};
43
44extern struct kmem_cache *cachefiles_object_jar;
45
46/*
47 * Cache files cache definition
48 */
49struct cachefiles_cache {
50 struct fscache_cache cache; /* FS-Cache record */
51 struct vfsmount *mnt; /* mountpoint holding the cache */
52 struct dentry *graveyard; /* directory into which dead objects go */
53 struct file *cachefilesd; /* manager daemon handle */
54 const struct cred *cache_cred; /* security override for accessing cache */
55 struct mutex daemon_mutex; /* command serialisation mutex */
56 wait_queue_head_t daemon_pollwq; /* poll waitqueue for daemon */
57 struct rb_root active_nodes; /* active nodes (can't be culled) */
58 rwlock_t active_lock; /* lock for active_nodes */
59 atomic_t gravecounter; /* graveyard uniquifier */
60 unsigned frun_percent; /* when to stop culling (% files) */
61 unsigned fcull_percent; /* when to start culling (% files) */
62 unsigned fstop_percent; /* when to stop allocating (% files) */
63 unsigned brun_percent; /* when to stop culling (% blocks) */
64 unsigned bcull_percent; /* when to start culling (% blocks) */
65 unsigned bstop_percent; /* when to stop allocating (% blocks) */
66 unsigned bsize; /* cache's block size */
67 unsigned bshift; /* min(ilog2(PAGE_SIZE / bsize), 0) */
68 uint64_t frun; /* when to stop culling */
69 uint64_t fcull; /* when to start culling */
70 uint64_t fstop; /* when to stop allocating */
71 sector_t brun; /* when to stop culling */
72 sector_t bcull; /* when to start culling */
73 sector_t bstop; /* when to stop allocating */
74 unsigned long flags;
75#define CACHEFILES_READY 0 /* T if cache prepared */
76#define CACHEFILES_DEAD 1 /* T if cache dead */
77#define CACHEFILES_CULLING 2 /* T if cull engaged */
78#define CACHEFILES_STATE_CHANGED 3 /* T if state changed (poll trigger) */
79 char *rootdirname; /* name of cache root directory */
80 char *secctx; /* LSM security context */
81 char *tag; /* cache binding tag */
82};
83
84/*
85 * backing file read tracking
86 */
87struct cachefiles_one_read {
88 wait_queue_t monitor; /* link into monitored waitqueue */
89 struct page *back_page; /* backing file page we're waiting for */
90 struct page *netfs_page; /* netfs page we're going to fill */
91 struct fscache_retrieval *op; /* retrieval op covering this */
92 struct list_head op_link; /* link in op's todo list */
93};
94
95/*
96 * backing file write tracking
97 */
98struct cachefiles_one_write {
99 struct page *netfs_page; /* netfs page to copy */
100 struct cachefiles_object *object;
101 struct list_head obj_link; /* link in object's lists */
102 fscache_rw_complete_t end_io_func;
103 void *context;
104};
105
106/*
107 * auxiliary data xattr buffer
108 */
109struct cachefiles_xattr {
110 uint16_t len;
111 uint8_t type;
112 uint8_t data[];
113};
114
115/*
116 * note change of state for daemon
117 */
118static inline void cachefiles_state_changed(struct cachefiles_cache *cache)
119{
120 set_bit(CACHEFILES_STATE_CHANGED, &cache->flags);
121 wake_up_all(&cache->daemon_pollwq);
122}
123
124/*
125 * cf-bind.c
126 */
127extern int cachefiles_daemon_bind(struct cachefiles_cache *cache, char *args);
128extern void cachefiles_daemon_unbind(struct cachefiles_cache *cache);
129
130/*
131 * cf-daemon.c
132 */
133extern const struct file_operations cachefiles_daemon_fops;
134
135extern int cachefiles_has_space(struct cachefiles_cache *cache,
136 unsigned fnr, unsigned bnr);
137
138/*
139 * cf-interface.c
140 */
141extern const struct fscache_cache_ops cachefiles_cache_ops;
142
143/*
144 * cf-key.c
145 */
146extern char *cachefiles_cook_key(const u8 *raw, int keylen, uint8_t type);
147
148/*
149 * cf-namei.c
150 */
151extern int cachefiles_delete_object(struct cachefiles_cache *cache,
152 struct cachefiles_object *object);
153extern int cachefiles_walk_to_object(struct cachefiles_object *parent,
154 struct cachefiles_object *object,
155 const char *key,
156 struct cachefiles_xattr *auxdata);
157extern struct dentry *cachefiles_get_directory(struct cachefiles_cache *cache,
158 struct dentry *dir,
159 const char *name);
160
161extern int cachefiles_cull(struct cachefiles_cache *cache, struct dentry *dir,
162 char *filename);
163
164extern int cachefiles_check_in_use(struct cachefiles_cache *cache,
165 struct dentry *dir, char *filename);
166
167/*
168 * cf-proc.c
169 */
170#ifdef CONFIG_CACHEFILES_HISTOGRAM
171extern atomic_t cachefiles_lookup_histogram[HZ];
172extern atomic_t cachefiles_mkdir_histogram[HZ];
173extern atomic_t cachefiles_create_histogram[HZ];
174
175extern int __init cachefiles_proc_init(void);
176extern void cachefiles_proc_cleanup(void);
177static inline
178void cachefiles_hist(atomic_t histogram[], unsigned long start_jif)
179{
180 unsigned long jif = jiffies - start_jif;
181 if (jif >= HZ)
182 jif = HZ - 1;
183 atomic_inc(&histogram[jif]);
184}
185
186#else
187#define cachefiles_proc_init() (0)
188#define cachefiles_proc_cleanup() do {} while (0)
189#define cachefiles_hist(hist, start_jif) do {} while (0)
190#endif
191
192/*
193 * cf-rdwr.c
194 */
195extern int cachefiles_read_or_alloc_page(struct fscache_retrieval *,
196 struct page *, gfp_t);
197extern int cachefiles_read_or_alloc_pages(struct fscache_retrieval *,
198 struct list_head *, unsigned *,
199 gfp_t);
200extern int cachefiles_allocate_page(struct fscache_retrieval *, struct page *,
201 gfp_t);
202extern int cachefiles_allocate_pages(struct fscache_retrieval *,
203 struct list_head *, unsigned *, gfp_t);
204extern int cachefiles_write_page(struct fscache_storage *, struct page *);
205extern void cachefiles_uncache_page(struct fscache_object *, struct page *);
206
207/*
208 * cf-security.c
209 */
210extern int cachefiles_get_security_ID(struct cachefiles_cache *cache);
211extern int cachefiles_determine_cache_security(struct cachefiles_cache *cache,
212 struct dentry *root,
213 const struct cred **_saved_cred);
214
215static inline void cachefiles_begin_secure(struct cachefiles_cache *cache,
216 const struct cred **_saved_cred)
217{
218 *_saved_cred = override_creds(cache->cache_cred);
219}
220
221static inline void cachefiles_end_secure(struct cachefiles_cache *cache,
222 const struct cred *saved_cred)
223{
224 revert_creds(saved_cred);
225}
226
227/*
228 * cf-xattr.c
229 */
230extern int cachefiles_check_object_type(struct cachefiles_object *object);
231extern int cachefiles_set_object_xattr(struct cachefiles_object *object,
232 struct cachefiles_xattr *auxdata);
233extern int cachefiles_update_object_xattr(struct cachefiles_object *object,
234 struct cachefiles_xattr *auxdata);
235extern int cachefiles_check_object_xattr(struct cachefiles_object *object,
236 struct cachefiles_xattr *auxdata);
237extern int cachefiles_remove_object_xattr(struct cachefiles_cache *cache,
238 struct dentry *dentry);
239
240
241/*
242 * error handling
243 */
244#define kerror(FMT, ...) printk(KERN_ERR "CacheFiles: "FMT"\n", ##__VA_ARGS__)
245
246#define cachefiles_io_error(___cache, FMT, ...) \
247do { \
248 kerror("I/O Error: " FMT, ##__VA_ARGS__); \
249 fscache_io_error(&(___cache)->cache); \
250 set_bit(CACHEFILES_DEAD, &(___cache)->flags); \
251} while (0)
252
253#define cachefiles_io_error_obj(object, FMT, ...) \
254do { \
255 struct cachefiles_cache *___cache; \
256 \
257 ___cache = container_of((object)->fscache.cache, \
258 struct cachefiles_cache, cache); \
259 cachefiles_io_error(___cache, FMT, ##__VA_ARGS__); \
260} while (0)
261
262
263/*
264 * debug tracing
265 */
266#define dbgprintk(FMT, ...) \
267 printk(KERN_DEBUG "[%-6.6s] "FMT"\n", current->comm, ##__VA_ARGS__)
268
269/* make sure we maintain the format strings, even when debugging is disabled */
270static inline void _dbprintk(const char *fmt, ...)
271 __attribute__((format(printf, 1, 2)));
272static inline void _dbprintk(const char *fmt, ...)
273{
274}
275
276#define kenter(FMT, ...) dbgprintk("==> %s("FMT")", __func__, ##__VA_ARGS__)
277#define kleave(FMT, ...) dbgprintk("<== %s()"FMT"", __func__, ##__VA_ARGS__)
278#define kdebug(FMT, ...) dbgprintk(FMT, ##__VA_ARGS__)
279
280
281#if defined(__KDEBUG)
282#define _enter(FMT, ...) kenter(FMT, ##__VA_ARGS__)
283#define _leave(FMT, ...) kleave(FMT, ##__VA_ARGS__)
284#define _debug(FMT, ...) kdebug(FMT, ##__VA_ARGS__)
285
286#elif defined(CONFIG_CACHEFILES_DEBUG)
287#define _enter(FMT, ...) \
288do { \
289 if (cachefiles_debug & CACHEFILES_DEBUG_KENTER) \
290 kenter(FMT, ##__VA_ARGS__); \
291} while (0)
292
293#define _leave(FMT, ...) \
294do { \
295 if (cachefiles_debug & CACHEFILES_DEBUG_KLEAVE) \
296 kleave(FMT, ##__VA_ARGS__); \
297} while (0)
298
299#define _debug(FMT, ...) \
300do { \
301 if (cachefiles_debug & CACHEFILES_DEBUG_KDEBUG) \
302 kdebug(FMT, ##__VA_ARGS__); \
303} while (0)
304
305#else
306#define _enter(FMT, ...) _dbprintk("==> %s("FMT")", __func__, ##__VA_ARGS__)
307#define _leave(FMT, ...) _dbprintk("<== %s()"FMT"", __func__, ##__VA_ARGS__)
308#define _debug(FMT, ...) _dbprintk(FMT, ##__VA_ARGS__)
309#endif
310
311#if 1 /* defined(__KDEBUGALL) */
312
313#define ASSERT(X) \
314do { \
315 if (unlikely(!(X))) { \
316 printk(KERN_ERR "\n"); \
317 printk(KERN_ERR "CacheFiles: Assertion failed\n"); \
318 BUG(); \
319 } \
320} while (0)
321
322#define ASSERTCMP(X, OP, Y) \
323do { \
324 if (unlikely(!((X) OP (Y)))) { \
325 printk(KERN_ERR "\n"); \
326 printk(KERN_ERR "CacheFiles: Assertion failed\n"); \
327 printk(KERN_ERR "%lx " #OP " %lx is false\n", \
328 (unsigned long)(X), (unsigned long)(Y)); \
329 BUG(); \
330 } \
331} while (0)
332
333#define ASSERTIF(C, X) \
334do { \
335 if (unlikely((C) && !(X))) { \
336 printk(KERN_ERR "\n"); \
337 printk(KERN_ERR "CacheFiles: Assertion failed\n"); \
338 BUG(); \
339 } \
340} while (0)
341
342#define ASSERTIFCMP(C, X, OP, Y) \
343do { \
344 if (unlikely((C) && !((X) OP (Y)))) { \
345 printk(KERN_ERR "\n"); \
346 printk(KERN_ERR "CacheFiles: Assertion failed\n"); \
347 printk(KERN_ERR "%lx " #OP " %lx is false\n", \
348 (unsigned long)(X), (unsigned long)(Y)); \
349 BUG(); \
350 } \
351} while (0)
352
353#else
354
355#define ASSERT(X) do {} while (0)
356#define ASSERTCMP(X, OP, Y) do {} while (0)
357#define ASSERTIF(C, X) do {} while (0)
358#define ASSERTIFCMP(C, X, OP, Y) do {} while (0)
359
360#endif
diff --git a/fs/cachefiles/key.c b/fs/cachefiles/key.c
new file mode 100644
index 000000000000..81b8b2b3a674
--- /dev/null
+++ b/fs/cachefiles/key.c
@@ -0,0 +1,159 @@
1/* Key to pathname encoder
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/slab.h>
13#include "internal.h"
14
15static const char cachefiles_charmap[64] =
16 "0123456789" /* 0 - 9 */
17 "abcdefghijklmnopqrstuvwxyz" /* 10 - 35 */
18 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" /* 36 - 61 */
19 "_-" /* 62 - 63 */
20 ;
21
22static const char cachefiles_filecharmap[256] = {
23 /* we skip space and tab and control chars */
24 [33 ... 46] = 1, /* '!' -> '.' */
25 /* we skip '/' as it's significant to pathwalk */
26 [48 ... 127] = 1, /* '0' -> '~' */
27};
28
29/*
30 * turn the raw key into something cooked
31 * - the raw key should include the length in the two bytes at the front
32 * - the key may be up to 514 bytes in length (including the length word)
33 * - "base64" encode the strange keys, mapping 3 bytes of raw to four of
34 * cooked
35 * - need to cut the cooked key into 252 char lengths (189 raw bytes)
36 */
37char *cachefiles_cook_key(const u8 *raw, int keylen, uint8_t type)
38{
39 unsigned char csum, ch;
40 unsigned int acc;
41 char *key;
42 int loop, len, max, seg, mark, print;
43
44 _enter(",%d", keylen);
45
46 BUG_ON(keylen < 2 || keylen > 514);
47
48 csum = raw[0] + raw[1];
49 print = 1;
50 for (loop = 2; loop < keylen; loop++) {
51 ch = raw[loop];
52 csum += ch;
53 print &= cachefiles_filecharmap[ch];
54 }
55
56 if (print) {
57 /* if the path is usable ASCII, then we render it directly */
58 max = keylen - 2;
59 max += 2; /* two base64'd length chars on the front */
60 max += 5; /* @checksum/M */
61 max += 3 * 2; /* maximum number of segment dividers (".../M")
62 * is ((514 + 251) / 252) = 3
63 */
64 max += 1; /* NUL on end */
65 } else {
66 /* calculate the maximum length of the cooked key */
67 keylen = (keylen + 2) / 3;
68
69 max = keylen * 4;
70 max += 5; /* @checksum/M */
71 max += 3 * 2; /* maximum number of segment dividers (".../M")
72 * is ((514 + 188) / 189) = 3
73 */
74 max += 1; /* NUL on end */
75 }
76
77 max += 1; /* 2nd NUL on end */
78
79 _debug("max: %d", max);
80
81 key = kmalloc(max, GFP_KERNEL);
82 if (!key)
83 return NULL;
84
85 len = 0;
86
87 /* build the cooked key */
88 sprintf(key, "@%02x%c+", (unsigned) csum, 0);
89 len = 5;
90 mark = len - 1;
91
92 if (print) {
93 acc = *(uint16_t *) raw;
94 raw += 2;
95
96 key[len + 1] = cachefiles_charmap[acc & 63];
97 acc >>= 6;
98 key[len] = cachefiles_charmap[acc & 63];
99 len += 2;
100
101 seg = 250;
102 for (loop = keylen; loop > 0; loop--) {
103 if (seg <= 0) {
104 key[len++] = '\0';
105 mark = len;
106 key[len++] = '+';
107 seg = 252;
108 }
109
110 key[len++] = *raw++;
111 ASSERT(len < max);
112 }
113
114 switch (type) {
115 case FSCACHE_COOKIE_TYPE_INDEX: type = 'I'; break;
116 case FSCACHE_COOKIE_TYPE_DATAFILE: type = 'D'; break;
117 default: type = 'S'; break;
118 }
119 } else {
120 seg = 252;
121 for (loop = keylen; loop > 0; loop--) {
122 if (seg <= 0) {
123 key[len++] = '\0';
124 mark = len;
125 key[len++] = '+';
126 seg = 252;
127 }
128
129 acc = *raw++;
130 acc |= *raw++ << 8;
131 acc |= *raw++ << 16;
132
133 _debug("acc: %06x", acc);
134
135 key[len++] = cachefiles_charmap[acc & 63];
136 acc >>= 6;
137 key[len++] = cachefiles_charmap[acc & 63];
138 acc >>= 6;
139 key[len++] = cachefiles_charmap[acc & 63];
140 acc >>= 6;
141 key[len++] = cachefiles_charmap[acc & 63];
142
143 ASSERT(len < max);
144 }
145
146 switch (type) {
147 case FSCACHE_COOKIE_TYPE_INDEX: type = 'J'; break;
148 case FSCACHE_COOKIE_TYPE_DATAFILE: type = 'E'; break;
149 default: type = 'T'; break;
150 }
151 }
152
153 key[mark] = type;
154 key[len++] = 0;
155 key[len] = 0;
156
157 _leave(" = %p %d", key, len);
158 return key;
159}
diff --git a/fs/cachefiles/main.c b/fs/cachefiles/main.c
new file mode 100644
index 000000000000..4bfa8cf43bf5
--- /dev/null
+++ b/fs/cachefiles/main.c
@@ -0,0 +1,106 @@
1/* Network filesystem caching backend to use cache files on a premounted
2 * filesystem
3 *
4 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public Licence
9 * as published by the Free Software Foundation; either version
10 * 2 of the Licence, or (at your option) any later version.
11 */
12
13#include <linux/module.h>
14#include <linux/init.h>
15#include <linux/sched.h>
16#include <linux/completion.h>
17#include <linux/slab.h>
18#include <linux/fs.h>
19#include <linux/file.h>
20#include <linux/namei.h>
21#include <linux/mount.h>
22#include <linux/statfs.h>
23#include <linux/sysctl.h>
24#include <linux/miscdevice.h>
25#include "internal.h"
26
27unsigned cachefiles_debug;
28module_param_named(debug, cachefiles_debug, uint, S_IWUSR | S_IRUGO);
29MODULE_PARM_DESC(cachefiles_debug, "CacheFiles debugging mask");
30
31MODULE_DESCRIPTION("Mounted-filesystem based cache");
32MODULE_AUTHOR("Red Hat, Inc.");
33MODULE_LICENSE("GPL");
34
35struct kmem_cache *cachefiles_object_jar;
36
37static struct miscdevice cachefiles_dev = {
38 .minor = MISC_DYNAMIC_MINOR,
39 .name = "cachefiles",
40 .fops = &cachefiles_daemon_fops,
41};
42
43static void cachefiles_object_init_once(void *_object)
44{
45 struct cachefiles_object *object = _object;
46
47 memset(object, 0, sizeof(*object));
48 spin_lock_init(&object->work_lock);
49}
50
51/*
52 * initialise the fs caching module
53 */
54static int __init cachefiles_init(void)
55{
56 int ret;
57
58 ret = misc_register(&cachefiles_dev);
59 if (ret < 0)
60 goto error_dev;
61
62 /* create an object jar */
63 ret = -ENOMEM;
64 cachefiles_object_jar =
65 kmem_cache_create("cachefiles_object_jar",
66 sizeof(struct cachefiles_object),
67 0,
68 SLAB_HWCACHE_ALIGN,
69 cachefiles_object_init_once);
70 if (!cachefiles_object_jar) {
71 printk(KERN_NOTICE
72 "CacheFiles: Failed to allocate an object jar\n");
73 goto error_object_jar;
74 }
75
76 ret = cachefiles_proc_init();
77 if (ret < 0)
78 goto error_proc;
79
80 printk(KERN_INFO "CacheFiles: Loaded\n");
81 return 0;
82
83error_proc:
84 kmem_cache_destroy(cachefiles_object_jar);
85error_object_jar:
86 misc_deregister(&cachefiles_dev);
87error_dev:
88 kerror("failed to register: %d", ret);
89 return ret;
90}
91
92fs_initcall(cachefiles_init);
93
94/*
95 * clean up on module removal
96 */
97static void __exit cachefiles_exit(void)
98{
99 printk(KERN_INFO "CacheFiles: Unloading\n");
100
101 cachefiles_proc_cleanup();
102 kmem_cache_destroy(cachefiles_object_jar);
103 misc_deregister(&cachefiles_dev);
104}
105
106module_exit(cachefiles_exit);
diff --git a/fs/cachefiles/namei.c b/fs/cachefiles/namei.c
new file mode 100644
index 000000000000..4ce818ae39ea
--- /dev/null
+++ b/fs/cachefiles/namei.c
@@ -0,0 +1,771 @@
1/* CacheFiles path walking and related routines
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/module.h>
13#include <linux/sched.h>
14#include <linux/file.h>
15#include <linux/fs.h>
16#include <linux/fsnotify.h>
17#include <linux/quotaops.h>
18#include <linux/xattr.h>
19#include <linux/mount.h>
20#include <linux/namei.h>
21#include <linux/security.h>
22#include "internal.h"
23
24static int cachefiles_wait_bit(void *flags)
25{
26 schedule();
27 return 0;
28}
29
30/*
31 * record the fact that an object is now active
32 */
33static void cachefiles_mark_object_active(struct cachefiles_cache *cache,
34 struct cachefiles_object *object)
35{
36 struct cachefiles_object *xobject;
37 struct rb_node **_p, *_parent = NULL;
38 struct dentry *dentry;
39
40 _enter(",%p", object);
41
42try_again:
43 write_lock(&cache->active_lock);
44
45 if (test_and_set_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags))
46 BUG();
47
48 dentry = object->dentry;
49 _p = &cache->active_nodes.rb_node;
50 while (*_p) {
51 _parent = *_p;
52 xobject = rb_entry(_parent,
53 struct cachefiles_object, active_node);
54
55 ASSERT(xobject != object);
56
57 if (xobject->dentry > dentry)
58 _p = &(*_p)->rb_left;
59 else if (xobject->dentry < dentry)
60 _p = &(*_p)->rb_right;
61 else
62 goto wait_for_old_object;
63 }
64
65 rb_link_node(&object->active_node, _parent, _p);
66 rb_insert_color(&object->active_node, &cache->active_nodes);
67
68 write_unlock(&cache->active_lock);
69 _leave("");
70 return;
71
72 /* an old object from a previous incarnation is hogging the slot - we
73 * need to wait for it to be destroyed */
74wait_for_old_object:
75 if (xobject->fscache.state < FSCACHE_OBJECT_DYING) {
76 printk(KERN_ERR "\n");
77 printk(KERN_ERR "CacheFiles: Error:"
78 " Unexpected object collision\n");
79 printk(KERN_ERR "xobject: OBJ%x\n",
80 xobject->fscache.debug_id);
81 printk(KERN_ERR "xobjstate=%s\n",
82 fscache_object_states[xobject->fscache.state]);
83 printk(KERN_ERR "xobjflags=%lx\n", xobject->fscache.flags);
84 printk(KERN_ERR "xobjevent=%lx [%lx]\n",
85 xobject->fscache.events, xobject->fscache.event_mask);
86 printk(KERN_ERR "xops=%u inp=%u exc=%u\n",
87 xobject->fscache.n_ops, xobject->fscache.n_in_progress,
88 xobject->fscache.n_exclusive);
89 printk(KERN_ERR "xcookie=%p [pr=%p nd=%p fl=%lx]\n",
90 xobject->fscache.cookie,
91 xobject->fscache.cookie->parent,
92 xobject->fscache.cookie->netfs_data,
93 xobject->fscache.cookie->flags);
94 printk(KERN_ERR "xparent=%p\n",
95 xobject->fscache.parent);
96 printk(KERN_ERR "object: OBJ%x\n",
97 object->fscache.debug_id);
98 printk(KERN_ERR "cookie=%p [pr=%p nd=%p fl=%lx]\n",
99 object->fscache.cookie,
100 object->fscache.cookie->parent,
101 object->fscache.cookie->netfs_data,
102 object->fscache.cookie->flags);
103 printk(KERN_ERR "parent=%p\n",
104 object->fscache.parent);
105 BUG();
106 }
107 atomic_inc(&xobject->usage);
108 write_unlock(&cache->active_lock);
109
110 _debug(">>> wait");
111 wait_on_bit(&xobject->flags, CACHEFILES_OBJECT_ACTIVE,
112 cachefiles_wait_bit, TASK_UNINTERRUPTIBLE);
113 _debug("<<< waited");
114
115 cache->cache.ops->put_object(&xobject->fscache);
116 goto try_again;
117}
118
119/*
120 * delete an object representation from the cache
121 * - file backed objects are unlinked
122 * - directory backed objects are stuffed into the graveyard for userspace to
123 * delete
124 * - unlocks the directory mutex
125 */
126static int cachefiles_bury_object(struct cachefiles_cache *cache,
127 struct dentry *dir,
128 struct dentry *rep)
129{
130 struct dentry *grave, *trap;
131 char nbuffer[8 + 8 + 1];
132 int ret;
133
134 _enter(",'%*.*s','%*.*s'",
135 dir->d_name.len, dir->d_name.len, dir->d_name.name,
136 rep->d_name.len, rep->d_name.len, rep->d_name.name);
137
138 /* non-directories can just be unlinked */
139 if (!S_ISDIR(rep->d_inode->i_mode)) {
140 _debug("unlink stale object");
141 ret = vfs_unlink(dir->d_inode, rep);
142
143 mutex_unlock(&dir->d_inode->i_mutex);
144
145 if (ret == -EIO)
146 cachefiles_io_error(cache, "Unlink failed");
147
148 _leave(" = %d", ret);
149 return ret;
150 }
151
152 /* directories have to be moved to the graveyard */
153 _debug("move stale object to graveyard");
154 mutex_unlock(&dir->d_inode->i_mutex);
155
156try_again:
157 /* first step is to make up a grave dentry in the graveyard */
158 sprintf(nbuffer, "%08x%08x",
159 (uint32_t) get_seconds(),
160 (uint32_t) atomic_inc_return(&cache->gravecounter));
161
162 /* do the multiway lock magic */
163 trap = lock_rename(cache->graveyard, dir);
164
165 /* do some checks before getting the grave dentry */
166 if (rep->d_parent != dir) {
167 /* the entry was probably culled when we dropped the parent dir
168 * lock */
169 unlock_rename(cache->graveyard, dir);
170 _leave(" = 0 [culled?]");
171 return 0;
172 }
173
174 if (!S_ISDIR(cache->graveyard->d_inode->i_mode)) {
175 unlock_rename(cache->graveyard, dir);
176 cachefiles_io_error(cache, "Graveyard no longer a directory");
177 return -EIO;
178 }
179
180 if (trap == rep) {
181 unlock_rename(cache->graveyard, dir);
182 cachefiles_io_error(cache, "May not make directory loop");
183 return -EIO;
184 }
185
186 if (d_mountpoint(rep)) {
187 unlock_rename(cache->graveyard, dir);
188 cachefiles_io_error(cache, "Mountpoint in cache");
189 return -EIO;
190 }
191
192 grave = lookup_one_len(nbuffer, cache->graveyard, strlen(nbuffer));
193 if (IS_ERR(grave)) {
194 unlock_rename(cache->graveyard, dir);
195
196 if (PTR_ERR(grave) == -ENOMEM) {
197 _leave(" = -ENOMEM");
198 return -ENOMEM;
199 }
200
201 cachefiles_io_error(cache, "Lookup error %ld",
202 PTR_ERR(grave));
203 return -EIO;
204 }
205
206 if (grave->d_inode) {
207 unlock_rename(cache->graveyard, dir);
208 dput(grave);
209 grave = NULL;
210 cond_resched();
211 goto try_again;
212 }
213
214 if (d_mountpoint(grave)) {
215 unlock_rename(cache->graveyard, dir);
216 dput(grave);
217 cachefiles_io_error(cache, "Mountpoint in graveyard");
218 return -EIO;
219 }
220
221 /* target should not be an ancestor of source */
222 if (trap == grave) {
223 unlock_rename(cache->graveyard, dir);
224 dput(grave);
225 cachefiles_io_error(cache, "May not make directory loop");
226 return -EIO;
227 }
228
229 /* attempt the rename */
230 ret = vfs_rename(dir->d_inode, rep, cache->graveyard->d_inode, grave);
231 if (ret != 0 && ret != -ENOMEM)
232 cachefiles_io_error(cache, "Rename failed with error %d", ret);
233
234 unlock_rename(cache->graveyard, dir);
235 dput(grave);
236 _leave(" = 0");
237 return 0;
238}
239
240/*
241 * delete an object representation from the cache
242 */
243int cachefiles_delete_object(struct cachefiles_cache *cache,
244 struct cachefiles_object *object)
245{
246 struct dentry *dir;
247 int ret;
248
249 _enter(",{%p}", object->dentry);
250
251 ASSERT(object->dentry);
252 ASSERT(object->dentry->d_inode);
253 ASSERT(object->dentry->d_parent);
254
255 dir = dget_parent(object->dentry);
256
257 mutex_lock(&dir->d_inode->i_mutex);
258 ret = cachefiles_bury_object(cache, dir, object->dentry);
259
260 dput(dir);
261 _leave(" = %d", ret);
262 return ret;
263}
264
265/*
266 * walk from the parent object to the child object through the backing
267 * filesystem, creating directories as we go
268 */
269int cachefiles_walk_to_object(struct cachefiles_object *parent,
270 struct cachefiles_object *object,
271 const char *key,
272 struct cachefiles_xattr *auxdata)
273{
274 struct cachefiles_cache *cache;
275 struct dentry *dir, *next = NULL;
276 unsigned long start;
277 const char *name;
278 int ret, nlen;
279
280 _enter("{%p},,%s,", parent->dentry, key);
281
282 cache = container_of(parent->fscache.cache,
283 struct cachefiles_cache, cache);
284
285 ASSERT(parent->dentry);
286 ASSERT(parent->dentry->d_inode);
287
288 if (!(S_ISDIR(parent->dentry->d_inode->i_mode))) {
289 // TODO: convert file to dir
290 _leave("looking up in none directory");
291 return -ENOBUFS;
292 }
293
294 dir = dget(parent->dentry);
295
296advance:
297 /* attempt to transit the first directory component */
298 name = key;
299 nlen = strlen(key);
300
301 /* key ends in a double NUL */
302 key = key + nlen + 1;
303 if (!*key)
304 key = NULL;
305
306lookup_again:
307 /* search the current directory for the element name */
308 _debug("lookup '%s'", name);
309
310 mutex_lock(&dir->d_inode->i_mutex);
311
312 start = jiffies;
313 next = lookup_one_len(name, dir, nlen);
314 cachefiles_hist(cachefiles_lookup_histogram, start);
315 if (IS_ERR(next))
316 goto lookup_error;
317
318 _debug("next -> %p %s", next, next->d_inode ? "positive" : "negative");
319
320 if (!key)
321 object->new = !next->d_inode;
322
323 /* if this element of the path doesn't exist, then the lookup phase
324 * failed, and we can release any readers in the certain knowledge that
325 * there's nothing for them to actually read */
326 if (!next->d_inode)
327 fscache_object_lookup_negative(&object->fscache);
328
329 /* we need to create the object if it's negative */
330 if (key || object->type == FSCACHE_COOKIE_TYPE_INDEX) {
331 /* index objects and intervening tree levels must be subdirs */
332 if (!next->d_inode) {
333 ret = cachefiles_has_space(cache, 1, 0);
334 if (ret < 0)
335 goto create_error;
336
337 start = jiffies;
338 ret = vfs_mkdir(dir->d_inode, next, 0);
339 cachefiles_hist(cachefiles_mkdir_histogram, start);
340 if (ret < 0)
341 goto create_error;
342
343 ASSERT(next->d_inode);
344
345 _debug("mkdir -> %p{%p{ino=%lu}}",
346 next, next->d_inode, next->d_inode->i_ino);
347
348 } else if (!S_ISDIR(next->d_inode->i_mode)) {
349 kerror("inode %lu is not a directory",
350 next->d_inode->i_ino);
351 ret = -ENOBUFS;
352 goto error;
353 }
354
355 } else {
356 /* non-index objects start out life as files */
357 if (!next->d_inode) {
358 ret = cachefiles_has_space(cache, 1, 0);
359 if (ret < 0)
360 goto create_error;
361
362 start = jiffies;
363 ret = vfs_create(dir->d_inode, next, S_IFREG, NULL);
364 cachefiles_hist(cachefiles_create_histogram, start);
365 if (ret < 0)
366 goto create_error;
367
368 ASSERT(next->d_inode);
369
370 _debug("create -> %p{%p{ino=%lu}}",
371 next, next->d_inode, next->d_inode->i_ino);
372
373 } else if (!S_ISDIR(next->d_inode->i_mode) &&
374 !S_ISREG(next->d_inode->i_mode)
375 ) {
376 kerror("inode %lu is not a file or directory",
377 next->d_inode->i_ino);
378 ret = -ENOBUFS;
379 goto error;
380 }
381 }
382
383 /* process the next component */
384 if (key) {
385 _debug("advance");
386 mutex_unlock(&dir->d_inode->i_mutex);
387 dput(dir);
388 dir = next;
389 next = NULL;
390 goto advance;
391 }
392
393 /* we've found the object we were looking for */
394 object->dentry = next;
395
396 /* if we've found that the terminal object exists, then we need to
397 * check its attributes and delete it if it's out of date */
398 if (!object->new) {
399 _debug("validate '%*.*s'",
400 next->d_name.len, next->d_name.len, next->d_name.name);
401
402 ret = cachefiles_check_object_xattr(object, auxdata);
403 if (ret == -ESTALE) {
404 /* delete the object (the deleter drops the directory
405 * mutex) */
406 object->dentry = NULL;
407
408 ret = cachefiles_bury_object(cache, dir, next);
409 dput(next);
410 next = NULL;
411
412 if (ret < 0)
413 goto delete_error;
414
415 _debug("redo lookup");
416 goto lookup_again;
417 }
418 }
419
420 /* note that we're now using this object */
421 cachefiles_mark_object_active(cache, object);
422
423 mutex_unlock(&dir->d_inode->i_mutex);
424 dput(dir);
425 dir = NULL;
426
427 _debug("=== OBTAINED_OBJECT ===");
428
429 if (object->new) {
430 /* attach data to a newly constructed terminal object */
431 ret = cachefiles_set_object_xattr(object, auxdata);
432 if (ret < 0)
433 goto check_error;
434 } else {
435 /* always update the atime on an object we've just looked up
436 * (this is used to keep track of culling, and atimes are only
437 * updated by read, write and readdir but not lookup or
438 * open) */
439 touch_atime(cache->mnt, next);
440 }
441
442 /* open a file interface onto a data file */
443 if (object->type != FSCACHE_COOKIE_TYPE_INDEX) {
444 if (S_ISREG(object->dentry->d_inode->i_mode)) {
445 const struct address_space_operations *aops;
446
447 ret = -EPERM;
448 aops = object->dentry->d_inode->i_mapping->a_ops;
449 if (!aops->bmap)
450 goto check_error;
451
452 object->backer = object->dentry;
453 } else {
454 BUG(); // TODO: open file in data-class subdir
455 }
456 }
457
458 object->new = 0;
459 fscache_obtained_object(&object->fscache);
460
461 _leave(" = 0 [%lu]", object->dentry->d_inode->i_ino);
462 return 0;
463
464create_error:
465 _debug("create error %d", ret);
466 if (ret == -EIO)
467 cachefiles_io_error(cache, "Create/mkdir failed");
468 goto error;
469
470check_error:
471 _debug("check error %d", ret);
472 write_lock(&cache->active_lock);
473 rb_erase(&object->active_node, &cache->active_nodes);
474 clear_bit(CACHEFILES_OBJECT_ACTIVE, &object->flags);
475 wake_up_bit(&object->flags, CACHEFILES_OBJECT_ACTIVE);
476 write_unlock(&cache->active_lock);
477
478 dput(object->dentry);
479 object->dentry = NULL;
480 goto error_out;
481
482delete_error:
483 _debug("delete error %d", ret);
484 goto error_out2;
485
486lookup_error:
487 _debug("lookup error %ld", PTR_ERR(next));
488 ret = PTR_ERR(next);
489 if (ret == -EIO)
490 cachefiles_io_error(cache, "Lookup failed");
491 next = NULL;
492error:
493 mutex_unlock(&dir->d_inode->i_mutex);
494 dput(next);
495error_out2:
496 dput(dir);
497error_out:
498 if (ret == -ENOSPC)
499 ret = -ENOBUFS;
500
501 _leave(" = error %d", -ret);
502 return ret;
503}
504
505/*
506 * get a subdirectory
507 */
508struct dentry *cachefiles_get_directory(struct cachefiles_cache *cache,
509 struct dentry *dir,
510 const char *dirname)
511{
512 struct dentry *subdir;
513 unsigned long start;
514 int ret;
515
516 _enter(",,%s", dirname);
517
518 /* search the current directory for the element name */
519 mutex_lock(&dir->d_inode->i_mutex);
520
521 start = jiffies;
522 subdir = lookup_one_len(dirname, dir, strlen(dirname));
523 cachefiles_hist(cachefiles_lookup_histogram, start);
524 if (IS_ERR(subdir)) {
525 if (PTR_ERR(subdir) == -ENOMEM)
526 goto nomem_d_alloc;
527 goto lookup_error;
528 }
529
530 _debug("subdir -> %p %s",
531 subdir, subdir->d_inode ? "positive" : "negative");
532
533 /* we need to create the subdir if it doesn't exist yet */
534 if (!subdir->d_inode) {
535 ret = cachefiles_has_space(cache, 1, 0);
536 if (ret < 0)
537 goto mkdir_error;
538
539 _debug("attempt mkdir");
540
541 ret = vfs_mkdir(dir->d_inode, subdir, 0700);
542 if (ret < 0)
543 goto mkdir_error;
544
545 ASSERT(subdir->d_inode);
546
547 _debug("mkdir -> %p{%p{ino=%lu}}",
548 subdir,
549 subdir->d_inode,
550 subdir->d_inode->i_ino);
551 }
552
553 mutex_unlock(&dir->d_inode->i_mutex);
554
555 /* we need to make sure the subdir is a directory */
556 ASSERT(subdir->d_inode);
557
558 if (!S_ISDIR(subdir->d_inode->i_mode)) {
559 kerror("%s is not a directory", dirname);
560 ret = -EIO;
561 goto check_error;
562 }
563
564 ret = -EPERM;
565 if (!subdir->d_inode->i_op ||
566 !subdir->d_inode->i_op->setxattr ||
567 !subdir->d_inode->i_op->getxattr ||
568 !subdir->d_inode->i_op->lookup ||
569 !subdir->d_inode->i_op->mkdir ||
570 !subdir->d_inode->i_op->create ||
571 !subdir->d_inode->i_op->rename ||
572 !subdir->d_inode->i_op->rmdir ||
573 !subdir->d_inode->i_op->unlink)
574 goto check_error;
575
576 _leave(" = [%lu]", subdir->d_inode->i_ino);
577 return subdir;
578
579check_error:
580 dput(subdir);
581 _leave(" = %d [check]", ret);
582 return ERR_PTR(ret);
583
584mkdir_error:
585 mutex_unlock(&dir->d_inode->i_mutex);
586 dput(subdir);
587 kerror("mkdir %s failed with error %d", dirname, ret);
588 return ERR_PTR(ret);
589
590lookup_error:
591 mutex_unlock(&dir->d_inode->i_mutex);
592 ret = PTR_ERR(subdir);
593 kerror("Lookup %s failed with error %d", dirname, ret);
594 return ERR_PTR(ret);
595
596nomem_d_alloc:
597 mutex_unlock(&dir->d_inode->i_mutex);
598 _leave(" = -ENOMEM");
599 return ERR_PTR(-ENOMEM);
600}
601
602/*
603 * find out if an object is in use or not
604 * - if finds object and it's not in use:
605 * - returns a pointer to the object and a reference on it
606 * - returns with the directory locked
607 */
608static struct dentry *cachefiles_check_active(struct cachefiles_cache *cache,
609 struct dentry *dir,
610 char *filename)
611{
612 struct cachefiles_object *object;
613 struct rb_node *_n;
614 struct dentry *victim;
615 unsigned long start;
616 int ret;
617
618 //_enter(",%*.*s/,%s",
619 // dir->d_name.len, dir->d_name.len, dir->d_name.name, filename);
620
621 /* look up the victim */
622 mutex_lock_nested(&dir->d_inode->i_mutex, 1);
623
624 start = jiffies;
625 victim = lookup_one_len(filename, dir, strlen(filename));
626 cachefiles_hist(cachefiles_lookup_histogram, start);
627 if (IS_ERR(victim))
628 goto lookup_error;
629
630 //_debug("victim -> %p %s",
631 // victim, victim->d_inode ? "positive" : "negative");
632
633 /* if the object is no longer there then we probably retired the object
634 * at the netfs's request whilst the cull was in progress
635 */
636 if (!victim->d_inode) {
637 mutex_unlock(&dir->d_inode->i_mutex);
638 dput(victim);
639 _leave(" = -ENOENT [absent]");
640 return ERR_PTR(-ENOENT);
641 }
642
643 /* check to see if we're using this object */
644 read_lock(&cache->active_lock);
645
646 _n = cache->active_nodes.rb_node;
647
648 while (_n) {
649 object = rb_entry(_n, struct cachefiles_object, active_node);
650
651 if (object->dentry > victim)
652 _n = _n->rb_left;
653 else if (object->dentry < victim)
654 _n = _n->rb_right;
655 else
656 goto object_in_use;
657 }
658
659 read_unlock(&cache->active_lock);
660
661 //_leave(" = %p", victim);
662 return victim;
663
664object_in_use:
665 read_unlock(&cache->active_lock);
666 mutex_unlock(&dir->d_inode->i_mutex);
667 dput(victim);
668 //_leave(" = -EBUSY [in use]");
669 return ERR_PTR(-EBUSY);
670
671lookup_error:
672 mutex_unlock(&dir->d_inode->i_mutex);
673 ret = PTR_ERR(victim);
674 if (ret == -ENOENT) {
675 /* file or dir now absent - probably retired by netfs */
676 _leave(" = -ESTALE [absent]");
677 return ERR_PTR(-ESTALE);
678 }
679
680 if (ret == -EIO) {
681 cachefiles_io_error(cache, "Lookup failed");
682 } else if (ret != -ENOMEM) {
683 kerror("Internal error: %d", ret);
684 ret = -EIO;
685 }
686
687 _leave(" = %d", ret);
688 return ERR_PTR(ret);
689}
690
691/*
692 * cull an object if it's not in use
693 * - called only by cache manager daemon
694 */
695int cachefiles_cull(struct cachefiles_cache *cache, struct dentry *dir,
696 char *filename)
697{
698 struct dentry *victim;
699 int ret;
700
701 _enter(",%*.*s/,%s",
702 dir->d_name.len, dir->d_name.len, dir->d_name.name, filename);
703
704 victim = cachefiles_check_active(cache, dir, filename);
705 if (IS_ERR(victim))
706 return PTR_ERR(victim);
707
708 _debug("victim -> %p %s",
709 victim, victim->d_inode ? "positive" : "negative");
710
711 /* okay... the victim is not being used so we can cull it
712 * - start by marking it as stale
713 */
714 _debug("victim is cullable");
715
716 ret = cachefiles_remove_object_xattr(cache, victim);
717 if (ret < 0)
718 goto error_unlock;
719
720 /* actually remove the victim (drops the dir mutex) */
721 _debug("bury");
722
723 ret = cachefiles_bury_object(cache, dir, victim);
724 if (ret < 0)
725 goto error;
726
727 dput(victim);
728 _leave(" = 0");
729 return 0;
730
731error_unlock:
732 mutex_unlock(&dir->d_inode->i_mutex);
733error:
734 dput(victim);
735 if (ret == -ENOENT) {
736 /* file or dir now absent - probably retired by netfs */
737 _leave(" = -ESTALE [absent]");
738 return -ESTALE;
739 }
740
741 if (ret != -ENOMEM) {
742 kerror("Internal error: %d", ret);
743 ret = -EIO;
744 }
745
746 _leave(" = %d", ret);
747 return ret;
748}
749
750/*
751 * find out if an object is in use or not
752 * - called only by cache manager daemon
753 * - returns -EBUSY or 0 to indicate whether an object is in use or not
754 */
755int cachefiles_check_in_use(struct cachefiles_cache *cache, struct dentry *dir,
756 char *filename)
757{
758 struct dentry *victim;
759
760 //_enter(",%*.*s/,%s",
761 // dir->d_name.len, dir->d_name.len, dir->d_name.name, filename);
762
763 victim = cachefiles_check_active(cache, dir, filename);
764 if (IS_ERR(victim))
765 return PTR_ERR(victim);
766
767 mutex_unlock(&dir->d_inode->i_mutex);
768 dput(victim);
769 //_leave(" = 0");
770 return 0;
771}
diff --git a/fs/cachefiles/proc.c b/fs/cachefiles/proc.c
new file mode 100644
index 000000000000..eccd33941199
--- /dev/null
+++ b/fs/cachefiles/proc.c
@@ -0,0 +1,134 @@
1/* CacheFiles statistics
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/module.h>
13#include <linux/proc_fs.h>
14#include <linux/seq_file.h>
15#include "internal.h"
16
17atomic_t cachefiles_lookup_histogram[HZ];
18atomic_t cachefiles_mkdir_histogram[HZ];
19atomic_t cachefiles_create_histogram[HZ];
20
21/*
22 * display the latency histogram
23 */
24static int cachefiles_histogram_show(struct seq_file *m, void *v)
25{
26 unsigned long index;
27 unsigned x, y, z, t;
28
29 switch ((unsigned long) v) {
30 case 1:
31 seq_puts(m, "JIFS SECS LOOKUPS MKDIRS CREATES\n");
32 return 0;
33 case 2:
34 seq_puts(m, "===== ===== ========= ========= =========\n");
35 return 0;
36 default:
37 index = (unsigned long) v - 3;
38 x = atomic_read(&cachefiles_lookup_histogram[index]);
39 y = atomic_read(&cachefiles_mkdir_histogram[index]);
40 z = atomic_read(&cachefiles_create_histogram[index]);
41 if (x == 0 && y == 0 && z == 0)
42 return 0;
43
44 t = (index * 1000) / HZ;
45
46 seq_printf(m, "%4lu 0.%03u %9u %9u %9u\n", index, t, x, y, z);
47 return 0;
48 }
49}
50
51/*
52 * set up the iterator to start reading from the first line
53 */
54static void *cachefiles_histogram_start(struct seq_file *m, loff_t *_pos)
55{
56 if ((unsigned long long)*_pos >= HZ + 2)
57 return NULL;
58 if (*_pos == 0)
59 *_pos = 1;
60 return (void *)(unsigned long) *_pos;
61}
62
63/*
64 * move to the next line
65 */
66static void *cachefiles_histogram_next(struct seq_file *m, void *v, loff_t *pos)
67{
68 (*pos)++;
69 return (unsigned long long)*pos > HZ + 2 ?
70 NULL : (void *)(unsigned long) *pos;
71}
72
73/*
74 * clean up after reading
75 */
76static void cachefiles_histogram_stop(struct seq_file *m, void *v)
77{
78}
79
80static const struct seq_operations cachefiles_histogram_ops = {
81 .start = cachefiles_histogram_start,
82 .stop = cachefiles_histogram_stop,
83 .next = cachefiles_histogram_next,
84 .show = cachefiles_histogram_show,
85};
86
87/*
88 * open "/proc/fs/cachefiles/XXX" which provide statistics summaries
89 */
90static int cachefiles_histogram_open(struct inode *inode, struct file *file)
91{
92 return seq_open(file, &cachefiles_histogram_ops);
93}
94
95static const struct file_operations cachefiles_histogram_fops = {
96 .owner = THIS_MODULE,
97 .open = cachefiles_histogram_open,
98 .read = seq_read,
99 .llseek = seq_lseek,
100 .release = seq_release,
101};
102
103/*
104 * initialise the /proc/fs/cachefiles/ directory
105 */
106int __init cachefiles_proc_init(void)
107{
108 _enter("");
109
110 if (!proc_mkdir("fs/cachefiles", NULL))
111 goto error_dir;
112
113 if (!proc_create("fs/cachefiles/histogram", S_IFREG | 0444, NULL,
114 &cachefiles_histogram_fops))
115 goto error_histogram;
116
117 _leave(" = 0");
118 return 0;
119
120error_histogram:
121 remove_proc_entry("fs/cachefiles", NULL);
122error_dir:
123 _leave(" = -ENOMEM");
124 return -ENOMEM;
125}
126
127/*
128 * clean up the /proc/fs/cachefiles/ directory
129 */
130void cachefiles_proc_cleanup(void)
131{
132 remove_proc_entry("fs/cachefiles/histogram", NULL);
133 remove_proc_entry("fs/cachefiles", NULL);
134}
diff --git a/fs/cachefiles/rdwr.c b/fs/cachefiles/rdwr.c
new file mode 100644
index 000000000000..a69787e7dd96
--- /dev/null
+++ b/fs/cachefiles/rdwr.c
@@ -0,0 +1,879 @@
1/* Storage object read/write
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/mount.h>
13#include <linux/file.h>
14#include "internal.h"
15
16/*
17 * detect wake up events generated by the unlocking of pages in which we're
18 * interested
19 * - we use this to detect read completion of backing pages
20 * - the caller holds the waitqueue lock
21 */
22static int cachefiles_read_waiter(wait_queue_t *wait, unsigned mode,
23 int sync, void *_key)
24{
25 struct cachefiles_one_read *monitor =
26 container_of(wait, struct cachefiles_one_read, monitor);
27 struct cachefiles_object *object;
28 struct wait_bit_key *key = _key;
29 struct page *page = wait->private;
30
31 ASSERT(key);
32
33 _enter("{%lu},%u,%d,{%p,%u}",
34 monitor->netfs_page->index, mode, sync,
35 key->flags, key->bit_nr);
36
37 if (key->flags != &page->flags ||
38 key->bit_nr != PG_locked)
39 return 0;
40
41 _debug("--- monitor %p %lx ---", page, page->flags);
42
43 if (!PageUptodate(page) && !PageError(page))
44 dump_stack();
45
46 /* remove from the waitqueue */
47 list_del(&wait->task_list);
48
49 /* move onto the action list and queue for FS-Cache thread pool */
50 ASSERT(monitor->op);
51
52 object = container_of(monitor->op->op.object,
53 struct cachefiles_object, fscache);
54
55 spin_lock(&object->work_lock);
56 list_add_tail(&monitor->op_link, &monitor->op->to_do);
57 spin_unlock(&object->work_lock);
58
59 fscache_enqueue_retrieval(monitor->op);
60 return 0;
61}
62
63/*
64 * copy data from backing pages to netfs pages to complete a read operation
65 * - driven by FS-Cache's thread pool
66 */
67static void cachefiles_read_copier(struct fscache_operation *_op)
68{
69 struct cachefiles_one_read *monitor;
70 struct cachefiles_object *object;
71 struct fscache_retrieval *op;
72 struct pagevec pagevec;
73 int error, max;
74
75 op = container_of(_op, struct fscache_retrieval, op);
76 object = container_of(op->op.object,
77 struct cachefiles_object, fscache);
78
79 _enter("{ino=%lu}", object->backer->d_inode->i_ino);
80
81 pagevec_init(&pagevec, 0);
82
83 max = 8;
84 spin_lock_irq(&object->work_lock);
85
86 while (!list_empty(&op->to_do)) {
87 monitor = list_entry(op->to_do.next,
88 struct cachefiles_one_read, op_link);
89 list_del(&monitor->op_link);
90
91 spin_unlock_irq(&object->work_lock);
92
93 _debug("- copy {%lu}", monitor->back_page->index);
94
95 error = -EIO;
96 if (PageUptodate(monitor->back_page)) {
97 copy_highpage(monitor->netfs_page, monitor->back_page);
98
99 pagevec_add(&pagevec, monitor->netfs_page);
100 fscache_mark_pages_cached(monitor->op, &pagevec);
101 error = 0;
102 }
103
104 if (error)
105 cachefiles_io_error_obj(
106 object,
107 "Readpage failed on backing file %lx",
108 (unsigned long) monitor->back_page->flags);
109
110 page_cache_release(monitor->back_page);
111
112 fscache_end_io(op, monitor->netfs_page, error);
113 page_cache_release(monitor->netfs_page);
114 fscache_put_retrieval(op);
115 kfree(monitor);
116
117 /* let the thread pool have some air occasionally */
118 max--;
119 if (max < 0 || need_resched()) {
120 if (!list_empty(&op->to_do))
121 fscache_enqueue_retrieval(op);
122 _leave(" [maxed out]");
123 return;
124 }
125
126 spin_lock_irq(&object->work_lock);
127 }
128
129 spin_unlock_irq(&object->work_lock);
130 _leave("");
131}
132
133/*
134 * read the corresponding page to the given set from the backing file
135 * - an uncertain page is simply discarded, to be tried again another time
136 */
137static int cachefiles_read_backing_file_one(struct cachefiles_object *object,
138 struct fscache_retrieval *op,
139 struct page *netpage,
140 struct pagevec *pagevec)
141{
142 struct cachefiles_one_read *monitor;
143 struct address_space *bmapping;
144 struct page *newpage, *backpage;
145 int ret;
146
147 _enter("");
148
149 pagevec_reinit(pagevec);
150
151 _debug("read back %p{%lu,%d}",
152 netpage, netpage->index, page_count(netpage));
153
154 monitor = kzalloc(sizeof(*monitor), GFP_KERNEL);
155 if (!monitor)
156 goto nomem;
157
158 monitor->netfs_page = netpage;
159 monitor->op = fscache_get_retrieval(op);
160
161 init_waitqueue_func_entry(&monitor->monitor, cachefiles_read_waiter);
162
163 /* attempt to get hold of the backing page */
164 bmapping = object->backer->d_inode->i_mapping;
165 newpage = NULL;
166
167 for (;;) {
168 backpage = find_get_page(bmapping, netpage->index);
169 if (backpage)
170 goto backing_page_already_present;
171
172 if (!newpage) {
173 newpage = page_cache_alloc_cold(bmapping);
174 if (!newpage)
175 goto nomem_monitor;
176 }
177
178 ret = add_to_page_cache(newpage, bmapping,
179 netpage->index, GFP_KERNEL);
180 if (ret == 0)
181 goto installed_new_backing_page;
182 if (ret != -EEXIST)
183 goto nomem_page;
184 }
185
186 /* we've installed a new backing page, so now we need to add it
187 * to the LRU list and start it reading */
188installed_new_backing_page:
189 _debug("- new %p", newpage);
190
191 backpage = newpage;
192 newpage = NULL;
193
194 page_cache_get(backpage);
195 pagevec_add(pagevec, backpage);
196 __pagevec_lru_add_file(pagevec);
197
198read_backing_page:
199 ret = bmapping->a_ops->readpage(NULL, backpage);
200 if (ret < 0)
201 goto read_error;
202
203 /* set the monitor to transfer the data across */
204monitor_backing_page:
205 _debug("- monitor add");
206
207 /* install the monitor */
208 page_cache_get(monitor->netfs_page);
209 page_cache_get(backpage);
210 monitor->back_page = backpage;
211 monitor->monitor.private = backpage;
212 add_page_wait_queue(backpage, &monitor->monitor);
213 monitor = NULL;
214
215 /* but the page may have been read before the monitor was installed, so
216 * the monitor may miss the event - so we have to ensure that we do get
217 * one in such a case */
218 if (trylock_page(backpage)) {
219 _debug("jumpstart %p {%lx}", backpage, backpage->flags);
220 unlock_page(backpage);
221 }
222 goto success;
223
224 /* if the backing page is already present, it can be in one of
225 * three states: read in progress, read failed or read okay */
226backing_page_already_present:
227 _debug("- present");
228
229 if (newpage) {
230 page_cache_release(newpage);
231 newpage = NULL;
232 }
233
234 if (PageError(backpage))
235 goto io_error;
236
237 if (PageUptodate(backpage))
238 goto backing_page_already_uptodate;
239
240 if (!trylock_page(backpage))
241 goto monitor_backing_page;
242 _debug("read %p {%lx}", backpage, backpage->flags);
243 goto read_backing_page;
244
245 /* the backing page is already up to date, attach the netfs
246 * page to the pagecache and LRU and copy the data across */
247backing_page_already_uptodate:
248 _debug("- uptodate");
249
250 pagevec_add(pagevec, netpage);
251 fscache_mark_pages_cached(op, pagevec);
252
253 copy_highpage(netpage, backpage);
254 fscache_end_io(op, netpage, 0);
255
256success:
257 _debug("success");
258 ret = 0;
259
260out:
261 if (backpage)
262 page_cache_release(backpage);
263 if (monitor) {
264 fscache_put_retrieval(monitor->op);
265 kfree(monitor);
266 }
267 _leave(" = %d", ret);
268 return ret;
269
270read_error:
271 _debug("read error %d", ret);
272 if (ret == -ENOMEM)
273 goto out;
274io_error:
275 cachefiles_io_error_obj(object, "Page read error on backing file");
276 ret = -ENOBUFS;
277 goto out;
278
279nomem_page:
280 page_cache_release(newpage);
281nomem_monitor:
282 fscache_put_retrieval(monitor->op);
283 kfree(monitor);
284nomem:
285 _leave(" = -ENOMEM");
286 return -ENOMEM;
287}
288
289/*
290 * read a page from the cache or allocate a block in which to store it
291 * - cache withdrawal is prevented by the caller
292 * - returns -EINTR if interrupted
293 * - returns -ENOMEM if ran out of memory
294 * - returns -ENOBUFS if no buffers can be made available
295 * - returns -ENOBUFS if page is beyond EOF
296 * - if the page is backed by a block in the cache:
297 * - a read will be started which will call the callback on completion
298 * - 0 will be returned
299 * - else if the page is unbacked:
300 * - the metadata will be retained
301 * - -ENODATA will be returned
302 */
303int cachefiles_read_or_alloc_page(struct fscache_retrieval *op,
304 struct page *page,
305 gfp_t gfp)
306{
307 struct cachefiles_object *object;
308 struct cachefiles_cache *cache;
309 struct pagevec pagevec;
310 struct inode *inode;
311 sector_t block0, block;
312 unsigned shift;
313 int ret;
314
315 object = container_of(op->op.object,
316 struct cachefiles_object, fscache);
317 cache = container_of(object->fscache.cache,
318 struct cachefiles_cache, cache);
319
320 _enter("{%p},{%lx},,,", object, page->index);
321
322 if (!object->backer)
323 return -ENOBUFS;
324
325 inode = object->backer->d_inode;
326 ASSERT(S_ISREG(inode->i_mode));
327 ASSERT(inode->i_mapping->a_ops->bmap);
328 ASSERT(inode->i_mapping->a_ops->readpages);
329
330 /* calculate the shift required to use bmap */
331 if (inode->i_sb->s_blocksize > PAGE_SIZE)
332 return -ENOBUFS;
333
334 shift = PAGE_SHIFT - inode->i_sb->s_blocksize_bits;
335
336 op->op.flags = FSCACHE_OP_FAST;
337 op->op.processor = cachefiles_read_copier;
338
339 pagevec_init(&pagevec, 0);
340
341 /* we assume the absence or presence of the first block is a good
342 * enough indication for the page as a whole
343 * - TODO: don't use bmap() for this as it is _not_ actually good
344 * enough for this as it doesn't indicate errors, but it's all we've
345 * got for the moment
346 */
347 block0 = page->index;
348 block0 <<= shift;
349
350 block = inode->i_mapping->a_ops->bmap(inode->i_mapping, block0);
351 _debug("%llx -> %llx",
352 (unsigned long long) block0,
353 (unsigned long long) block);
354
355 if (block) {
356 /* submit the apparently valid page to the backing fs to be
357 * read from disk */
358 ret = cachefiles_read_backing_file_one(object, op, page,
359 &pagevec);
360 } else if (cachefiles_has_space(cache, 0, 1) == 0) {
361 /* there's space in the cache we can use */
362 pagevec_add(&pagevec, page);
363 fscache_mark_pages_cached(op, &pagevec);
364 ret = -ENODATA;
365 } else {
366 ret = -ENOBUFS;
367 }
368
369 _leave(" = %d", ret);
370 return ret;
371}
372
373/*
374 * read the corresponding pages to the given set from the backing file
375 * - any uncertain pages are simply discarded, to be tried again another time
376 */
377static int cachefiles_read_backing_file(struct cachefiles_object *object,
378 struct fscache_retrieval *op,
379 struct list_head *list,
380 struct pagevec *mark_pvec)
381{
382 struct cachefiles_one_read *monitor = NULL;
383 struct address_space *bmapping = object->backer->d_inode->i_mapping;
384 struct pagevec lru_pvec;
385 struct page *newpage = NULL, *netpage, *_n, *backpage = NULL;
386 int ret = 0;
387
388 _enter("");
389
390 pagevec_init(&lru_pvec, 0);
391
392 list_for_each_entry_safe(netpage, _n, list, lru) {
393 list_del(&netpage->lru);
394
395 _debug("read back %p{%lu,%d}",
396 netpage, netpage->index, page_count(netpage));
397
398 if (!monitor) {
399 monitor = kzalloc(sizeof(*monitor), GFP_KERNEL);
400 if (!monitor)
401 goto nomem;
402
403 monitor->op = fscache_get_retrieval(op);
404 init_waitqueue_func_entry(&monitor->monitor,
405 cachefiles_read_waiter);
406 }
407
408 for (;;) {
409 backpage = find_get_page(bmapping, netpage->index);
410 if (backpage)
411 goto backing_page_already_present;
412
413 if (!newpage) {
414 newpage = page_cache_alloc_cold(bmapping);
415 if (!newpage)
416 goto nomem;
417 }
418
419 ret = add_to_page_cache(newpage, bmapping,
420 netpage->index, GFP_KERNEL);
421 if (ret == 0)
422 goto installed_new_backing_page;
423 if (ret != -EEXIST)
424 goto nomem;
425 }
426
427 /* we've installed a new backing page, so now we need to add it
428 * to the LRU list and start it reading */
429 installed_new_backing_page:
430 _debug("- new %p", newpage);
431
432 backpage = newpage;
433 newpage = NULL;
434
435 page_cache_get(backpage);
436 if (!pagevec_add(&lru_pvec, backpage))
437 __pagevec_lru_add_file(&lru_pvec);
438
439 reread_backing_page:
440 ret = bmapping->a_ops->readpage(NULL, backpage);
441 if (ret < 0)
442 goto read_error;
443
444 /* add the netfs page to the pagecache and LRU, and set the
445 * monitor to transfer the data across */
446 monitor_backing_page:
447 _debug("- monitor add");
448
449 ret = add_to_page_cache(netpage, op->mapping, netpage->index,
450 GFP_KERNEL);
451 if (ret < 0) {
452 if (ret == -EEXIST) {
453 page_cache_release(netpage);
454 continue;
455 }
456 goto nomem;
457 }
458
459 page_cache_get(netpage);
460 if (!pagevec_add(&lru_pvec, netpage))
461 __pagevec_lru_add_file(&lru_pvec);
462
463 /* install a monitor */
464 page_cache_get(netpage);
465 monitor->netfs_page = netpage;
466
467 page_cache_get(backpage);
468 monitor->back_page = backpage;
469 monitor->monitor.private = backpage;
470 add_page_wait_queue(backpage, &monitor->monitor);
471 monitor = NULL;
472
473 /* but the page may have been read before the monitor was
474 * installed, so the monitor may miss the event - so we have to
475 * ensure that we do get one in such a case */
476 if (trylock_page(backpage)) {
477 _debug("2unlock %p {%lx}", backpage, backpage->flags);
478 unlock_page(backpage);
479 }
480
481 page_cache_release(backpage);
482 backpage = NULL;
483
484 page_cache_release(netpage);
485 netpage = NULL;
486 continue;
487
488 /* if the backing page is already present, it can be in one of
489 * three states: read in progress, read failed or read okay */
490 backing_page_already_present:
491 _debug("- present %p", backpage);
492
493 if (PageError(backpage))
494 goto io_error;
495
496 if (PageUptodate(backpage))
497 goto backing_page_already_uptodate;
498
499 _debug("- not ready %p{%lx}", backpage, backpage->flags);
500
501 if (!trylock_page(backpage))
502 goto monitor_backing_page;
503
504 if (PageError(backpage)) {
505 _debug("error %lx", backpage->flags);
506 unlock_page(backpage);
507 goto io_error;
508 }
509
510 if (PageUptodate(backpage))
511 goto backing_page_already_uptodate_unlock;
512
513 /* we've locked a page that's neither up to date nor erroneous,
514 * so we need to attempt to read it again */
515 goto reread_backing_page;
516
517 /* the backing page is already up to date, attach the netfs
518 * page to the pagecache and LRU and copy the data across */
519 backing_page_already_uptodate_unlock:
520 _debug("uptodate %lx", backpage->flags);
521 unlock_page(backpage);
522 backing_page_already_uptodate:
523 _debug("- uptodate");
524
525 ret = add_to_page_cache(netpage, op->mapping, netpage->index,
526 GFP_KERNEL);
527 if (ret < 0) {
528 if (ret == -EEXIST) {
529 page_cache_release(netpage);
530 continue;
531 }
532 goto nomem;
533 }
534
535 copy_highpage(netpage, backpage);
536
537 page_cache_release(backpage);
538 backpage = NULL;
539
540 if (!pagevec_add(mark_pvec, netpage))
541 fscache_mark_pages_cached(op, mark_pvec);
542
543 page_cache_get(netpage);
544 if (!pagevec_add(&lru_pvec, netpage))
545 __pagevec_lru_add_file(&lru_pvec);
546
547 fscache_end_io(op, netpage, 0);
548 page_cache_release(netpage);
549 netpage = NULL;
550 continue;
551 }
552
553 netpage = NULL;
554
555 _debug("out");
556
557out:
558 /* tidy up */
559 pagevec_lru_add_file(&lru_pvec);
560
561 if (newpage)
562 page_cache_release(newpage);
563 if (netpage)
564 page_cache_release(netpage);
565 if (backpage)
566 page_cache_release(backpage);
567 if (monitor) {
568 fscache_put_retrieval(op);
569 kfree(monitor);
570 }
571
572 list_for_each_entry_safe(netpage, _n, list, lru) {
573 list_del(&netpage->lru);
574 page_cache_release(netpage);
575 }
576
577 _leave(" = %d", ret);
578 return ret;
579
580nomem:
581 _debug("nomem");
582 ret = -ENOMEM;
583 goto out;
584
585read_error:
586 _debug("read error %d", ret);
587 if (ret == -ENOMEM)
588 goto out;
589io_error:
590 cachefiles_io_error_obj(object, "Page read error on backing file");
591 ret = -ENOBUFS;
592 goto out;
593}
594
595/*
596 * read a list of pages from the cache or allocate blocks in which to store
597 * them
598 */
599int cachefiles_read_or_alloc_pages(struct fscache_retrieval *op,
600 struct list_head *pages,
601 unsigned *nr_pages,
602 gfp_t gfp)
603{
604 struct cachefiles_object *object;
605 struct cachefiles_cache *cache;
606 struct list_head backpages;
607 struct pagevec pagevec;
608 struct inode *inode;
609 struct page *page, *_n;
610 unsigned shift, nrbackpages;
611 int ret, ret2, space;
612
613 object = container_of(op->op.object,
614 struct cachefiles_object, fscache);
615 cache = container_of(object->fscache.cache,
616 struct cachefiles_cache, cache);
617
618 _enter("{OBJ%x,%d},,%d,,",
619 object->fscache.debug_id, atomic_read(&op->op.usage),
620 *nr_pages);
621
622 if (!object->backer)
623 return -ENOBUFS;
624
625 space = 1;
626 if (cachefiles_has_space(cache, 0, *nr_pages) < 0)
627 space = 0;
628
629 inode = object->backer->d_inode;
630 ASSERT(S_ISREG(inode->i_mode));
631 ASSERT(inode->i_mapping->a_ops->bmap);
632 ASSERT(inode->i_mapping->a_ops->readpages);
633
634 /* calculate the shift required to use bmap */
635 if (inode->i_sb->s_blocksize > PAGE_SIZE)
636 return -ENOBUFS;
637
638 shift = PAGE_SHIFT - inode->i_sb->s_blocksize_bits;
639
640 pagevec_init(&pagevec, 0);
641
642 op->op.flags = FSCACHE_OP_FAST;
643 op->op.processor = cachefiles_read_copier;
644
645 INIT_LIST_HEAD(&backpages);
646 nrbackpages = 0;
647
648 ret = space ? -ENODATA : -ENOBUFS;
649 list_for_each_entry_safe(page, _n, pages, lru) {
650 sector_t block0, block;
651
652 /* we assume the absence or presence of the first block is a
653 * good enough indication for the page as a whole
654 * - TODO: don't use bmap() for this as it is _not_ actually
655 * good enough for this as it doesn't indicate errors, but
656 * it's all we've got for the moment
657 */
658 block0 = page->index;
659 block0 <<= shift;
660
661 block = inode->i_mapping->a_ops->bmap(inode->i_mapping,
662 block0);
663 _debug("%llx -> %llx",
664 (unsigned long long) block0,
665 (unsigned long long) block);
666
667 if (block) {
668 /* we have data - add it to the list to give to the
669 * backing fs */
670 list_move(&page->lru, &backpages);
671 (*nr_pages)--;
672 nrbackpages++;
673 } else if (space && pagevec_add(&pagevec, page) == 0) {
674 fscache_mark_pages_cached(op, &pagevec);
675 ret = -ENODATA;
676 }
677 }
678
679 if (pagevec_count(&pagevec) > 0)
680 fscache_mark_pages_cached(op, &pagevec);
681
682 if (list_empty(pages))
683 ret = 0;
684
685 /* submit the apparently valid pages to the backing fs to be read from
686 * disk */
687 if (nrbackpages > 0) {
688 ret2 = cachefiles_read_backing_file(object, op, &backpages,
689 &pagevec);
690 if (ret2 == -ENOMEM || ret2 == -EINTR)
691 ret = ret2;
692 }
693
694 if (pagevec_count(&pagevec) > 0)
695 fscache_mark_pages_cached(op, &pagevec);
696
697 _leave(" = %d [nr=%u%s]",
698 ret, *nr_pages, list_empty(pages) ? " empty" : "");
699 return ret;
700}
701
702/*
703 * allocate a block in the cache in which to store a page
704 * - cache withdrawal is prevented by the caller
705 * - returns -EINTR if interrupted
706 * - returns -ENOMEM if ran out of memory
707 * - returns -ENOBUFS if no buffers can be made available
708 * - returns -ENOBUFS if page is beyond EOF
709 * - otherwise:
710 * - the metadata will be retained
711 * - 0 will be returned
712 */
713int cachefiles_allocate_page(struct fscache_retrieval *op,
714 struct page *page,
715 gfp_t gfp)
716{
717 struct cachefiles_object *object;
718 struct cachefiles_cache *cache;
719 struct pagevec pagevec;
720 int ret;
721
722 object = container_of(op->op.object,
723 struct cachefiles_object, fscache);
724 cache = container_of(object->fscache.cache,
725 struct cachefiles_cache, cache);
726
727 _enter("%p,{%lx},", object, page->index);
728
729 ret = cachefiles_has_space(cache, 0, 1);
730 if (ret == 0) {
731 pagevec_init(&pagevec, 0);
732 pagevec_add(&pagevec, page);
733 fscache_mark_pages_cached(op, &pagevec);
734 } else {
735 ret = -ENOBUFS;
736 }
737
738 _leave(" = %d", ret);
739 return ret;
740}
741
742/*
743 * allocate blocks in the cache in which to store a set of pages
744 * - cache withdrawal is prevented by the caller
745 * - returns -EINTR if interrupted
746 * - returns -ENOMEM if ran out of memory
747 * - returns -ENOBUFS if some buffers couldn't be made available
748 * - returns -ENOBUFS if some pages are beyond EOF
749 * - otherwise:
750 * - -ENODATA will be returned
751 * - metadata will be retained for any page marked
752 */
753int cachefiles_allocate_pages(struct fscache_retrieval *op,
754 struct list_head *pages,
755 unsigned *nr_pages,
756 gfp_t gfp)
757{
758 struct cachefiles_object *object;
759 struct cachefiles_cache *cache;
760 struct pagevec pagevec;
761 struct page *page;
762 int ret;
763
764 object = container_of(op->op.object,
765 struct cachefiles_object, fscache);
766 cache = container_of(object->fscache.cache,
767 struct cachefiles_cache, cache);
768
769 _enter("%p,,,%d,", object, *nr_pages);
770
771 ret = cachefiles_has_space(cache, 0, *nr_pages);
772 if (ret == 0) {
773 pagevec_init(&pagevec, 0);
774
775 list_for_each_entry(page, pages, lru) {
776 if (pagevec_add(&pagevec, page) == 0)
777 fscache_mark_pages_cached(op, &pagevec);
778 }
779
780 if (pagevec_count(&pagevec) > 0)
781 fscache_mark_pages_cached(op, &pagevec);
782 ret = -ENODATA;
783 } else {
784 ret = -ENOBUFS;
785 }
786
787 _leave(" = %d", ret);
788 return ret;
789}
790
791/*
792 * request a page be stored in the cache
793 * - cache withdrawal is prevented by the caller
794 * - this request may be ignored if there's no cache block available, in which
795 * case -ENOBUFS will be returned
796 * - if the op is in progress, 0 will be returned
797 */
798int cachefiles_write_page(struct fscache_storage *op, struct page *page)
799{
800 struct cachefiles_object *object;
801 struct cachefiles_cache *cache;
802 mm_segment_t old_fs;
803 struct file *file;
804 loff_t pos;
805 void *data;
806 int ret;
807
808 ASSERT(op != NULL);
809 ASSERT(page != NULL);
810
811 object = container_of(op->op.object,
812 struct cachefiles_object, fscache);
813
814 _enter("%p,%p{%lx},,,", object, page, page->index);
815
816 if (!object->backer) {
817 _leave(" = -ENOBUFS");
818 return -ENOBUFS;
819 }
820
821 ASSERT(S_ISREG(object->backer->d_inode->i_mode));
822
823 cache = container_of(object->fscache.cache,
824 struct cachefiles_cache, cache);
825
826 /* write the page to the backing filesystem and let it store it in its
827 * own time */
828 dget(object->backer);
829 mntget(cache->mnt);
830 file = dentry_open(object->backer, cache->mnt, O_RDWR,
831 cache->cache_cred);
832 if (IS_ERR(file)) {
833 ret = PTR_ERR(file);
834 } else {
835 ret = -EIO;
836 if (file->f_op->write) {
837 pos = (loff_t) page->index << PAGE_SHIFT;
838 data = kmap(page);
839 old_fs = get_fs();
840 set_fs(KERNEL_DS);
841 ret = file->f_op->write(
842 file, (const void __user *) data, PAGE_SIZE,
843 &pos);
844 set_fs(old_fs);
845 kunmap(page);
846 if (ret != PAGE_SIZE)
847 ret = -EIO;
848 }
849 fput(file);
850 }
851
852 if (ret < 0) {
853 if (ret == -EIO)
854 cachefiles_io_error_obj(
855 object, "Write page to backing file failed");
856 ret = -ENOBUFS;
857 }
858
859 _leave(" = %d", ret);
860 return ret;
861}
862
863/*
864 * detach a backing block from a page
865 * - cache withdrawal is prevented by the caller
866 */
867void cachefiles_uncache_page(struct fscache_object *_object, struct page *page)
868{
869 struct cachefiles_object *object;
870 struct cachefiles_cache *cache;
871
872 object = container_of(_object, struct cachefiles_object, fscache);
873 cache = container_of(object->fscache.cache,
874 struct cachefiles_cache, cache);
875
876 _enter("%p,{%lu}", object, page->index);
877
878 spin_unlock(&object->fscache.cookie->lock);
879}
diff --git a/fs/cachefiles/security.c b/fs/cachefiles/security.c
new file mode 100644
index 000000000000..b5808cdb2232
--- /dev/null
+++ b/fs/cachefiles/security.c
@@ -0,0 +1,116 @@
1/* CacheFiles security management
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/fs.h>
13#include <linux/cred.h>
14#include "internal.h"
15
16/*
17 * determine the security context within which we access the cache from within
18 * the kernel
19 */
20int cachefiles_get_security_ID(struct cachefiles_cache *cache)
21{
22 struct cred *new;
23 int ret;
24
25 _enter("{%s}", cache->secctx);
26
27 new = prepare_kernel_cred(current);
28 if (!new) {
29 ret = -ENOMEM;
30 goto error;
31 }
32
33 if (cache->secctx) {
34 ret = set_security_override_from_ctx(new, cache->secctx);
35 if (ret < 0) {
36 put_cred(new);
37 printk(KERN_ERR "CacheFiles:"
38 " Security denies permission to nominate"
39 " security context: error %d\n",
40 ret);
41 goto error;
42 }
43 }
44
45 cache->cache_cred = new;
46 ret = 0;
47error:
48 _leave(" = %d", ret);
49 return ret;
50}
51
52/*
53 * see if mkdir and create can be performed in the root directory
54 */
55static int cachefiles_check_cache_dir(struct cachefiles_cache *cache,
56 struct dentry *root)
57{
58 int ret;
59
60 ret = security_inode_mkdir(root->d_inode, root, 0);
61 if (ret < 0) {
62 printk(KERN_ERR "CacheFiles:"
63 " Security denies permission to make dirs: error %d",
64 ret);
65 return ret;
66 }
67
68 ret = security_inode_create(root->d_inode, root, 0);
69 if (ret < 0)
70 printk(KERN_ERR "CacheFiles:"
71 " Security denies permission to create files: error %d",
72 ret);
73
74 return ret;
75}
76
77/*
78 * check the security details of the on-disk cache
79 * - must be called with security override in force
80 */
81int cachefiles_determine_cache_security(struct cachefiles_cache *cache,
82 struct dentry *root,
83 const struct cred **_saved_cred)
84{
85 struct cred *new;
86 int ret;
87
88 _enter("");
89
90 /* duplicate the cache creds for COW (the override is currently in
91 * force, so we can use prepare_creds() to do this) */
92 new = prepare_creds();
93 if (!new)
94 return -ENOMEM;
95
96 cachefiles_end_secure(cache, *_saved_cred);
97
98 /* use the cache root dir's security context as the basis with
99 * which create files */
100 ret = set_create_files_as(new, root->d_inode);
101 if (ret < 0) {
102 _leave(" = %d [cfa]", ret);
103 return ret;
104 }
105
106 put_cred(cache->cache_cred);
107 cache->cache_cred = new;
108
109 cachefiles_begin_secure(cache, _saved_cred);
110 ret = cachefiles_check_cache_dir(cache, root);
111
112 if (ret == -EOPNOTSUPP)
113 ret = 0;
114 _leave(" = %d", ret);
115 return ret;
116}
diff --git a/fs/cachefiles/xattr.c b/fs/cachefiles/xattr.c
new file mode 100644
index 000000000000..f3e7a0bf068b
--- /dev/null
+++ b/fs/cachefiles/xattr.c
@@ -0,0 +1,291 @@
1/* CacheFiles extended attribute management
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11
12#include <linux/module.h>
13#include <linux/sched.h>
14#include <linux/file.h>
15#include <linux/fs.h>
16#include <linux/fsnotify.h>
17#include <linux/quotaops.h>
18#include <linux/xattr.h>
19#include "internal.h"
20
21static const char cachefiles_xattr_cache[] =
22 XATTR_USER_PREFIX "CacheFiles.cache";
23
24/*
25 * check the type label on an object
26 * - done using xattrs
27 */
28int cachefiles_check_object_type(struct cachefiles_object *object)
29{
30 struct dentry *dentry = object->dentry;
31 char type[3], xtype[3];
32 int ret;
33
34 ASSERT(dentry);
35 ASSERT(dentry->d_inode);
36
37 if (!object->fscache.cookie)
38 strcpy(type, "C3");
39 else
40 snprintf(type, 3, "%02x", object->fscache.cookie->def->type);
41
42 _enter("%p{%s}", object, type);
43
44 /* attempt to install a type label directly */
45 ret = vfs_setxattr(dentry, cachefiles_xattr_cache, type, 2,
46 XATTR_CREATE);
47 if (ret == 0) {
48 _debug("SET"); /* we succeeded */
49 goto error;
50 }
51
52 if (ret != -EEXIST) {
53 kerror("Can't set xattr on %*.*s [%lu] (err %d)",
54 dentry->d_name.len, dentry->d_name.len,
55 dentry->d_name.name, dentry->d_inode->i_ino,
56 -ret);
57 goto error;
58 }
59
60 /* read the current type label */
61 ret = vfs_getxattr(dentry, cachefiles_xattr_cache, xtype, 3);
62 if (ret < 0) {
63 if (ret == -ERANGE)
64 goto bad_type_length;
65
66 kerror("Can't read xattr on %*.*s [%lu] (err %d)",
67 dentry->d_name.len, dentry->d_name.len,
68 dentry->d_name.name, dentry->d_inode->i_ino,
69 -ret);
70 goto error;
71 }
72
73 /* check the type is what we're expecting */
74 if (ret != 2)
75 goto bad_type_length;
76
77 if (xtype[0] != type[0] || xtype[1] != type[1])
78 goto bad_type;
79
80 ret = 0;
81
82error:
83 _leave(" = %d", ret);
84 return ret;
85
86bad_type_length:
87 kerror("Cache object %lu type xattr length incorrect",
88 dentry->d_inode->i_ino);
89 ret = -EIO;
90 goto error;
91
92bad_type:
93 xtype[2] = 0;
94 kerror("Cache object %*.*s [%lu] type %s not %s",
95 dentry->d_name.len, dentry->d_name.len,
96 dentry->d_name.name, dentry->d_inode->i_ino,
97 xtype, type);
98 ret = -EIO;
99 goto error;
100}
101
102/*
103 * set the state xattr on a cache file
104 */
105int cachefiles_set_object_xattr(struct cachefiles_object *object,
106 struct cachefiles_xattr *auxdata)
107{
108 struct dentry *dentry = object->dentry;
109 int ret;
110
111 ASSERT(object->fscache.cookie);
112 ASSERT(dentry);
113
114 _enter("%p,#%d", object, auxdata->len);
115
116 /* attempt to install the cache metadata directly */
117 _debug("SET %s #%u", object->fscache.cookie->def->name, auxdata->len);
118
119 ret = vfs_setxattr(dentry, cachefiles_xattr_cache,
120 &auxdata->type, auxdata->len,
121 XATTR_CREATE);
122 if (ret < 0 && ret != -ENOMEM)
123 cachefiles_io_error_obj(
124 object,
125 "Failed to set xattr with error %d", ret);
126
127 _leave(" = %d", ret);
128 return ret;
129}
130
131/*
132 * update the state xattr on a cache file
133 */
134int cachefiles_update_object_xattr(struct cachefiles_object *object,
135 struct cachefiles_xattr *auxdata)
136{
137 struct dentry *dentry = object->dentry;
138 int ret;
139
140 ASSERT(object->fscache.cookie);
141 ASSERT(dentry);
142
143 _enter("%p,#%d", object, auxdata->len);
144
145 /* attempt to install the cache metadata directly */
146 _debug("SET %s #%u", object->fscache.cookie->def->name, auxdata->len);
147
148 ret = vfs_setxattr(dentry, cachefiles_xattr_cache,
149 &auxdata->type, auxdata->len,
150 XATTR_REPLACE);
151 if (ret < 0 && ret != -ENOMEM)
152 cachefiles_io_error_obj(
153 object,
154 "Failed to update xattr with error %d", ret);
155
156 _leave(" = %d", ret);
157 return ret;
158}
159
160/*
161 * check the state xattr on a cache file
162 * - return -ESTALE if the object should be deleted
163 */
164int cachefiles_check_object_xattr(struct cachefiles_object *object,
165 struct cachefiles_xattr *auxdata)
166{
167 struct cachefiles_xattr *auxbuf;
168 struct dentry *dentry = object->dentry;
169 int ret;
170
171 _enter("%p,#%d", object, auxdata->len);
172
173 ASSERT(dentry);
174 ASSERT(dentry->d_inode);
175
176 auxbuf = kmalloc(sizeof(struct cachefiles_xattr) + 512, GFP_KERNEL);
177 if (!auxbuf) {
178 _leave(" = -ENOMEM");
179 return -ENOMEM;
180 }
181
182 /* read the current type label */
183 ret = vfs_getxattr(dentry, cachefiles_xattr_cache,
184 &auxbuf->type, 512 + 1);
185 if (ret < 0) {
186 if (ret == -ENODATA)
187 goto stale; /* no attribute - power went off
188 * mid-cull? */
189
190 if (ret == -ERANGE)
191 goto bad_type_length;
192
193 cachefiles_io_error_obj(object,
194 "Can't read xattr on %lu (err %d)",
195 dentry->d_inode->i_ino, -ret);
196 goto error;
197 }
198
199 /* check the on-disk object */
200 if (ret < 1)
201 goto bad_type_length;
202
203 if (auxbuf->type != auxdata->type)
204 goto stale;
205
206 auxbuf->len = ret;
207
208 /* consult the netfs */
209 if (object->fscache.cookie->def->check_aux) {
210 enum fscache_checkaux result;
211 unsigned int dlen;
212
213 dlen = auxbuf->len - 1;
214
215 _debug("checkaux %s #%u",
216 object->fscache.cookie->def->name, dlen);
217
218 result = fscache_check_aux(&object->fscache,
219 &auxbuf->data, dlen);
220
221 switch (result) {
222 /* entry okay as is */
223 case FSCACHE_CHECKAUX_OKAY:
224 goto okay;
225
226 /* entry requires update */
227 case FSCACHE_CHECKAUX_NEEDS_UPDATE:
228 break;
229
230 /* entry requires deletion */
231 case FSCACHE_CHECKAUX_OBSOLETE:
232 goto stale;
233
234 default:
235 BUG();
236 }
237
238 /* update the current label */
239 ret = vfs_setxattr(dentry, cachefiles_xattr_cache,
240 &auxdata->type, auxdata->len,
241 XATTR_REPLACE);
242 if (ret < 0) {
243 cachefiles_io_error_obj(object,
244 "Can't update xattr on %lu"
245 " (error %d)",
246 dentry->d_inode->i_ino, -ret);
247 goto error;
248 }
249 }
250
251okay:
252 ret = 0;
253
254error:
255 kfree(auxbuf);
256 _leave(" = %d", ret);
257 return ret;
258
259bad_type_length:
260 kerror("Cache object %lu xattr length incorrect",
261 dentry->d_inode->i_ino);
262 ret = -EIO;
263 goto error;
264
265stale:
266 ret = -ESTALE;
267 goto error;
268}
269
270/*
271 * remove the object's xattr to mark it stale
272 */
273int cachefiles_remove_object_xattr(struct cachefiles_cache *cache,
274 struct dentry *dentry)
275{
276 int ret;
277
278 ret = vfs_removexattr(dentry, cachefiles_xattr_cache);
279 if (ret < 0) {
280 if (ret == -ENOENT || ret == -ENODATA)
281 ret = 0;
282 else if (ret != -ENOMEM)
283 cachefiles_io_error(cache,
284 "Can't remove xattr from %lu"
285 " (error %d)",
286 dentry->d_inode->i_ino, -ret);
287 }
288
289 _leave(" = %d", ret);
290 return ret;
291}
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-23 03:28:29 -0500