Magellan Linux

Contents of /trunk/kernel26-magellan/patches-2.6.30-r1/0153-2.6.30-unionfs-2.5.1.patch

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1114 - (show annotations) (download)
Sun Aug 22 17:59:15 2010 UTC (13 years, 8 months ago) by niro
File size: 342115 byte(s)
-added

1 diff -Naur linux-2.6.30/Documentation/filesystems/00-INDEX linux-2.6.30-magellan/Documentation/filesystems/00-INDEX
2 --- linux-2.6.30/Documentation/filesystems/00-INDEX 2009-06-10 05:05:27.000000000 +0200
3 +++ linux-2.6.30-magellan/Documentation/filesystems/00-INDEX 2009-06-11 18:39:19.000000000 +0200
4 @@ -108,6 +108,8 @@
5 - info and mount options for the UDF filesystem.
6 ufs.txt
7 - info on the ufs filesystem.
8 +unionfs/
9 + - info on the unionfs filesystem
10 vfat.txt
11 - info on using the VFAT filesystem used in Windows NT and Windows 95
12 vfs.txt
13 diff -Naur linux-2.6.30/Documentation/filesystems/unionfs/00-INDEX linux-2.6.30-magellan/Documentation/filesystems/unionfs/00-INDEX
14 --- linux-2.6.30/Documentation/filesystems/unionfs/00-INDEX 1970-01-01 01:00:00.000000000 +0100
15 +++ linux-2.6.30-magellan/Documentation/filesystems/unionfs/00-INDEX 2009-06-11 18:39:19.000000000 +0200
16 @@ -0,0 +1,10 @@
17 +00-INDEX
18 + - this file.
19 +concepts.txt
20 + - A brief introduction of concepts.
21 +issues.txt
22 + - A summary of known issues with unionfs.
23 +rename.txt
24 + - Information regarding rename operations.
25 +usage.txt
26 + - Usage information and examples.
27 diff -Naur linux-2.6.30/Documentation/filesystems/unionfs/concepts.txt linux-2.6.30-magellan/Documentation/filesystems/unionfs/concepts.txt
28 --- linux-2.6.30/Documentation/filesystems/unionfs/concepts.txt 1970-01-01 01:00:00.000000000 +0100
29 +++ linux-2.6.30-magellan/Documentation/filesystems/unionfs/concepts.txt 2009-06-11 18:39:19.000000000 +0200
30 @@ -0,0 +1,287 @@
31 +Unionfs 2.x CONCEPTS:
32 +=====================
33 +
34 +This file describes the concepts needed by a namespace unification file
35 +system.
36 +
37 +
38 +Branch Priority:
39 +================
40 +
41 +Each branch is assigned a unique priority - starting from 0 (highest
42 +priority). No two branches can have the same priority.
43 +
44 +
45 +Branch Mode:
46 +============
47 +
48 +Each branch is assigned a mode - read-write or read-only. This allows
49 +directories on media mounted read-write to be used in a read-only manner.
50 +
51 +
52 +Whiteouts:
53 +==========
54 +
55 +A whiteout removes a file name from the namespace. Whiteouts are needed when
56 +one attempts to remove a file on a read-only branch.
57 +
58 +Suppose we have a two-branch union, where branch 0 is read-write and branch
59 +1 is read-only. And a file 'foo' on branch 1:
60 +
61 +./b0/
62 +./b1/
63 +./b1/foo
64 +
65 +The unified view would simply be:
66 +
67 +./union/
68 +./union/foo
69 +
70 +Since 'foo' is stored on a read-only branch, it cannot be removed. A
71 +whiteout is used to remove the name 'foo' from the unified namespace. Again,
72 +since branch 1 is read-only, the whiteout cannot be created there. So, we
73 +try on a higher priority (lower numerically) branch and create the whiteout
74 +there.
75 +
76 +./b0/
77 +./b0/.wh.foo
78 +./b1/
79 +./b1/foo
80 +
81 +Later, when Unionfs traverses branches (due to lookup or readdir), it
82 +eliminate 'foo' from the namespace (as well as the whiteout itself.)
83 +
84 +
85 +Opaque Directories:
86 +===================
87 +
88 +Assume we have a unionfs mount comprising of two branches. Branch 0 is
89 +empty; branch 1 has the directory /a and file /a/f. Let's say we mount a
90 +union of branch 0 as read-write and branch 1 as read-only. Now, let's say
91 +we try to perform the following operation in the union:
92 +
93 + rm -fr a
94 +
95 +Because branch 1 is not writable, we cannot physically remove the file /a/f
96 +or the directory /a. So instead, we will create a whiteout in branch 0
97 +named /.wh.a, masking out the name "a" from branch 1. Next, let's say we
98 +try to create a directory named "a" as follows:
99 +
100 + mkdir a
101 +
102 +Because we have a whiteout for "a" already, Unionfs behaves as if "a"
103 +doesn't exist, and thus will delete the whiteout and replace it with an
104 +actual directory named "a".
105 +
106 +The problem now is that if you try to "ls" in the union, Unionfs will
107 +perform is normal directory name unification, for *all* directories named
108 +"a" in all branches. This will cause the file /a/f from branch 1 to
109 +re-appear in the union's namespace, which violates Unix semantics.
110 +
111 +To avoid this problem, we have a different form of whiteouts for
112 +directories, called "opaque directories" (same as BSD Union Mount does).
113 +Whenever we replace a whiteout with a directory, that directory is marked as
114 +opaque. In Unionfs 2.x, it means that we create a file named
115 +/a/.wh.__dir_opaque in branch 0, after having created directory /a there.
116 +When unionfs notices that a directory is opaque, it stops all namespace
117 +operations (including merging readdir contents) at that opaque directory.
118 +This prevents re-exposing names from masked out directories.
119 +
120 +
121 +Duplicate Elimination:
122 +======================
123 +
124 +It is possible for files on different branches to have the same name.
125 +Unionfs then has to select which instance of the file to show to the user.
126 +Given the fact that each branch has a priority associated with it, the
127 +simplest solution is to take the instance from the highest priority
128 +(numerically lowest value) and "hide" the others.
129 +
130 +
131 +Unlinking:
132 +=========
133 +
134 +Unlink operation on non-directory instances is optimized to remove the
135 +maximum possible objects in case multiple underlying branches have the same
136 +file name. The unlink operation will first try to delete file instances
137 +from highest priority branch and then move further to delete from remaining
138 +branches in order of their decreasing priority. Consider a case (F..D..F),
139 +where F is a file and D is a directory of the same name; here, some
140 +intermediate branch could have an empty directory instance with the same
141 +name, so this operation also tries to delete this directory instance and
142 +proceed further to delete from next possible lower priority branch. The
143 +unionfs unlink operation will smoothly delete the files with same name from
144 +all possible underlying branches. In case if some error occurs, it creates
145 +whiteout in highest priority branch that will hide file instance in rest of
146 +the branches. An error could occur either if an unlink operations in any of
147 +the underlying branch failed or if a branch has no write permission.
148 +
149 +This unlinking policy is known as "delete all" and it has the benefit of
150 +overall reducing the number of inodes used by duplicate files, and further
151 +reducing the total number of inodes consumed by whiteouts. The cost is of
152 +extra processing, but testing shows this extra processing is well worth the
153 +savings.
154 +
155 +
156 +Copyup:
157 +=======
158 +
159 +When a change is made to the contents of a file's data or meta-data, they
160 +have to be stored somewhere. The best way is to create a copy of the
161 +original file on a branch that is writable, and then redirect the write
162 +though to this copy. The copy must be made on a higher priority branch so
163 +that lookup and readdir return this newer "version" of the file rather than
164 +the original (see duplicate elimination).
165 +
166 +An entire unionfs mount can be read-only or read-write. If it's read-only,
167 +then none of the branches will be written to, even if some of the branches
168 +are physically writeable. If the unionfs mount is read-write, then the
169 +leftmost (highest priority) branch must be writeable (for copyup to take
170 +place); the remaining branches can be any mix of read-write and read-only.
171 +
172 +In a writeable mount, unionfs will create new files/dir in the leftmost
173 +branch. If one tries to modify a file in a read-only branch/media, unionfs
174 +will copyup the file to the leftmost branch and modify it there. If you try
175 +to modify a file from a writeable branch which is not the leftmost branch,
176 +then unionfs will modify it in that branch; this is useful if you, say,
177 +unify differnet packages (e.g., apache, sendmail, ftpd, etc.) and you want
178 +changes to specific package files to remain logically in the directory where
179 +they came from.
180 +
181 +Cache Coherency:
182 +================
183 +
184 +Unionfs users often want to be able to modify files and directories directly
185 +on the lower branches, and have those changes be visible at the Unionfs
186 +level. This means that data (e.g., pages) and meta-data (dentries, inodes,
187 +open files, etc.) have to be synchronized between the upper and lower
188 +layers. In other words, the newest changes from a layer below have to be
189 +propagated to the Unionfs layer above. If the two layers are not in sync, a
190 +cache incoherency ensues, which could lead to application failures and even
191 +oopses. The Linux kernel, however, has a rather limited set of mechanisms
192 +to ensure this inter-layer cache coherency---so Unionfs has to do most of
193 +the hard work on its own.
194 +
195 +Maintaining Invariants:
196 +
197 +The way Unionfs ensures cache coherency is as follows. At each entry point
198 +to a Unionfs file system method, we call a utility function to validate the
199 +primary objects of this method. Generally, we call unionfs_file_revalidate
200 +on open files, and __unionfs_d_revalidate_chain on dentries (which also
201 +validates inodes). These utility functions check to see whether the upper
202 +Unionfs object is in sync with any of the lower objects that it represents.
203 +The checks we perform include whether the Unionfs superblock has a newer
204 +generation number, or if any of the lower objects mtime's or ctime's are
205 +newer. (Note: generation numbers change when branch-management commands are
206 +issued, so in a way, maintaining cache coherency is also very important for
207 +branch-management.) If indeed we determine that any Unionfs object is no
208 +longer in sync with its lower counterparts, then we rebuild that object
209 +similarly to how we do so for branch-management.
210 +
211 +While rebuilding Unionfs's objects, we also purge any page mappings and
212 +truncate inode pages (see fs/unionfs/dentry.c:purge_inode_data). This is to
213 +ensure that Unionfs will re-get the newer data from the lower branches. We
214 +perform this purging only if the Unionfs operation in question is a reading
215 +operation; if Unionfs is performing a data writing operation (e.g., ->write,
216 +->commit_write, etc.) then we do NOT flush the lower mappings/pages: this is
217 +because (1) a self-deadlock could occur and (2) the upper Unionfs pages are
218 +considered more authoritative anyway, as they are newer and will overwrite
219 +any lower pages.
220 +
221 +Unionfs maintains the following important invariant regarding mtime's,
222 +ctime's, and atime's: the upper inode object's times are the max() of all of
223 +the lower ones. For non-directory objects, there's only one object below,
224 +so the mapping is simple; for directory objects, there could me multiple
225 +lower objects and we have to sync up with the newest one of all the lower
226 +ones. This invariant is important to maintain, especially for directories
227 +(besides, we need this to be POSIX compliant). A union could comprise
228 +multiple writable branches, each of which could change. If we don't reflect
229 +the newest possible mtime/ctime, some applications could fail. For example,
230 +NFSv2/v3 exports check for newer directory mtimes on the server to determine
231 +if the client-side attribute cache should be purged.
232 +
233 +To maintain these important invariants, of course, Unionfs carefully
234 +synchronizes upper and lower times in various places. For example, if we
235 +copy-up a file to a top-level branch, the parent directory where the file
236 +was copied up to will now have a new mtime: so after a successful copy-up,
237 +we sync up with the new top-level branch's parent directory mtime.
238 +
239 +Implementation:
240 +
241 +This cache-coherency implementation is efficient because it defers any
242 +synchronizing between the upper and lower layers until absolutely needed.
243 +Consider the example a common situation where users perform a lot of lower
244 +changes, such as untarring a whole package. While these take place,
245 +typically the user doesn't access the files via Unionfs; only after the
246 +lower changes are done, does the user try to access the lower files. With
247 +our cache-coherency implementation, the entirety of the changes to the lower
248 +branches will not result in a single CPU cycle spent at the Unionfs level
249 +until the user invokes a system call that goes through Unionfs.
250 +
251 +We have considered two alternate cache-coherency designs. (1) Using the
252 +dentry/inode notify functionality to register interest in finding out about
253 +any lower changes. This is a somewhat limited and also a heavy-handed
254 +approach which could result in many notifications to the Unionfs layer upon
255 +each small change at the lower layer (imagine a file being modified multiple
256 +times in rapid succession). (2) Rewriting the VFS to support explicit
257 +callbacks from lower objects to upper objects. We began exploring such an
258 +implementation, but found it to be very complicated--it would have resulted
259 +in massive VFS/MM changes which are unlikely to be accepted by the LKML
260 +community. We therefore believe that our current cache-coherency design and
261 +implementation represent the best approach at this time.
262 +
263 +Limitations:
264 +
265 +Our implementation works in that as long as a user process will have caused
266 +Unionfs to be called, directly or indirectly, even to just do
267 +->d_revalidate; then we will have purged the current Unionfs data and the
268 +process will see the new data. For example, a process that continually
269 +re-reads the same file's data will see the NEW data as soon as the lower
270 +file had changed, upon the next read(2) syscall (even if the file is still
271 +open!) However, this doesn't work when the process re-reads the open file's
272 +data via mmap(2) (unless the user unmaps/closes the file and remaps/reopens
273 +it). Once we respond to ->readpage(s), then the kernel maps the page into
274 +the process's address space and there doesn't appear to be a way to force
275 +the kernel to invalidate those pages/mappings, and force the process to
276 +re-issue ->readpage. If there's a way to invalidate active mappings and
277 +force a ->readpage, let us know please (invalidate_inode_pages2 doesn't do
278 +the trick).
279 +
280 +Our current Unionfs code has to perform many file-revalidation calls. It
281 +would be really nice if the VFS would export an optional file system hook
282 +->file_revalidate (similarly to dentry->d_revalidate) that will be called
283 +before each VFS op that has a "struct file" in it.
284 +
285 +Certain file systems have micro-second granularity (or better) for inode
286 +times, and asynchronous actions could cause those times to change with some
287 +small delay. In such cases, Unionfs may see a changed inode time that only
288 +differs by a tiny fraction of a second: such a change may be a false
289 +positive indication that the lower object has changed, whereas if unionfs
290 +waits a little longer, that false indication will not be seen. (These false
291 +positives are harmless, because they would at most cause unionfs to
292 +re-validate an object that may need no revalidation, and print a debugging
293 +message that clutters the console/logs.) Therefore, to minimize the chances
294 +of these situations, we delay the detection of changed times by a small
295 +factor of a few seconds, called UNIONFS_MIN_CC_TIME (which defaults to 3
296 +seconds, as does NFS). This means that we will detect the change, only a
297 +couple of seconds later, if indeed the time change persists in the lower
298 +file object. This delayed detection has an added performance benefit: we
299 +reduce the number of times that unionfs has to revalidate objects, in case
300 +there's a lot of concurrent activity on both the upper and lower objects,
301 +for the same file(s). Lastly, this delayed time attribute detection is
302 +similar to how NFS clients operate (e.g., acregmin).
303 +
304 +Finally, there is no way currently in Linux to prevent lower directories
305 +from being moved around (i.e., topology changes); there's no way to prevent
306 +modifications to directory sub-trees of whole file systems which are mounted
307 +read-write. It is therefore possible for in-flight operations in unionfs to
308 +take place, while a lower directory is being moved around. Therefore, if
309 +you try to, say, create a new file in a directory through unionfs, while the
310 +directory is being moved around directly, then the new file may get created
311 +in the new location where that directory was moved to. This is a somewhat
312 +similar behaviour in NFS: an NFS client could be creating a new file while
313 +th NFS server is moving th directory around; the file will get successfully
314 +created in the new location. (The one exception in unionfs is that if the
315 +branch is marked read-only by unionfs, then a copyup will take place.)
316 +
317 +For more information, see <http://unionfs.filesystems.org/>.
318 diff -Naur linux-2.6.30/Documentation/filesystems/unionfs/issues.txt linux-2.6.30-magellan/Documentation/filesystems/unionfs/issues.txt
319 --- linux-2.6.30/Documentation/filesystems/unionfs/issues.txt 1970-01-01 01:00:00.000000000 +0100
320 +++ linux-2.6.30-magellan/Documentation/filesystems/unionfs/issues.txt 2009-06-11 18:39:19.000000000 +0200
321 @@ -0,0 +1,28 @@
322 +KNOWN Unionfs 2.x ISSUES:
323 +=========================
324 +
325 +1. Unionfs should not use lookup_one_len() on the underlying f/s as it
326 + confuses NFSv4. Currently, unionfs_lookup() passes lookup intents to the
327 + lower file-system, this eliminates part of the problem. The remaining
328 + calls to lookup_one_len may need to be changed to pass an intent. We are
329 + currently introducing VFS changes to fs/namei.c's do_path_lookup() to
330 + allow proper file lookup and opening in stackable file systems.
331 +
332 +2. Lockdep (a debugging feature) isn't aware of stacking, and so it
333 + incorrectly complains about locking problems. The problem boils down to
334 + this: Lockdep considers all objects of a certain type to be in the same
335 + class, for example, all inodes. Lockdep doesn't like to see a lock held
336 + on two inodes within the same task, and warns that it could lead to a
337 + deadlock. However, stackable file systems do precisely that: they lock
338 + an upper object, and then a lower object, in a strict order to avoid
339 + locking problems; in addition, Unionfs, as a fan-out file system, may
340 + have to lock several lower inodes. We are currently looking into Lockdep
341 + to see how to make it aware of stackable file systems. For now, we
342 + temporarily disable lockdep when calling vfs methods on lower objects,
343 + but only for those places where lockdep complained. While this solution
344 + may seem unclean, it is not without precedent: other places in the kernel
345 + also do similar temporary disabling, of course after carefully having
346 + checked that it is the right thing to do. Anyway, you get any warnings
347 + from Lockdep, please report them to the Unionfs maintainers.
348 +
349 +For more information, see <http://unionfs.filesystems.org/>.
350 diff -Naur linux-2.6.30/Documentation/filesystems/unionfs/rename.txt linux-2.6.30-magellan/Documentation/filesystems/unionfs/rename.txt
351 --- linux-2.6.30/Documentation/filesystems/unionfs/rename.txt 1970-01-01 01:00:00.000000000 +0100
352 +++ linux-2.6.30-magellan/Documentation/filesystems/unionfs/rename.txt 2009-06-11 18:39:19.000000000 +0200
353 @@ -0,0 +1,31 @@
354 +Rename is a complex beast. The following table shows which rename(2) operations
355 +should succeed and which should fail.
356 +
357 +o: success
358 +E: error (either unionfs or vfs)
359 +X: EXDEV
360 +
361 +none = file does not exist
362 +file = file is a file
363 +dir = file is a empty directory
364 +child= file is a non-empty directory
365 +wh = file is a directory containing only whiteouts; this makes it logically
366 + empty
367 +
368 + none file dir child wh
369 +file o o E E E
370 +dir o E o E o
371 +child X E X E X
372 +wh o E o E o
373 +
374 +
375 +Renaming directories:
376 +=====================
377 +
378 +Whenever a empty (either physically or logically) directory is being renamed,
379 +the following sequence of events should take place:
380 +
381 +1) Remove whiteouts from both source and destination directory
382 +2) Rename source to destination
383 +3) Make destination opaque to prevent anything under it from showing up
384 +
385 diff -Naur linux-2.6.30/Documentation/filesystems/unionfs/usage.txt linux-2.6.30-magellan/Documentation/filesystems/unionfs/usage.txt
386 --- linux-2.6.30/Documentation/filesystems/unionfs/usage.txt 1970-01-01 01:00:00.000000000 +0100
387 +++ linux-2.6.30-magellan/Documentation/filesystems/unionfs/usage.txt 2009-06-11 18:39:19.000000000 +0200
388 @@ -0,0 +1,134 @@
389 +Unionfs is a stackable unification file system, which can appear to merge
390 +the contents of several directories (branches), while keeping their physical
391 +content separate. Unionfs is useful for unified source tree management,
392 +merged contents of split CD-ROM, merged separate software package
393 +directories, data grids, and more. Unionfs allows any mix of read-only and
394 +read-write branches, as well as insertion and deletion of branches anywhere
395 +in the fan-out. To maintain Unix semantics, Unionfs handles elimination of
396 +duplicates, partial-error conditions, and more.
397 +
398 +GENERAL SYNTAX
399 +==============
400 +
401 +# mount -t unionfs -o <OPTIONS>,<BRANCH-OPTIONS> none MOUNTPOINT
402 +
403 +OPTIONS can be any legal combination of:
404 +
405 +- ro # mount file system read-only
406 +- rw # mount file system read-write
407 +- remount # remount the file system (see Branch Management below)
408 +- incgen # increment generation no. (see Cache Consistency below)
409 +
410 +BRANCH-OPTIONS can be either (1) a list of branches given to the "dirs="
411 +option, or (2) a list of individual branch manipulation commands, combined
412 +with the "remount" option, and is further described in the "Branch
413 +Management" section below.
414 +
415 +The syntax for the "dirs=" mount option is:
416 +
417 + dirs=branch[=ro|=rw][:...]
418 +
419 +The "dirs=" option takes a colon-delimited list of directories to compose
420 +the union, with an optional branch mode for each of those directories.
421 +Directories that come earlier (specified first, on the left) in the list
422 +have a higher precedence than those which come later. Additionally,
423 +read-only or read-write permissions of the branch can be specified by
424 +appending =ro or =rw (default) to each directory. See the Copyup section in
425 +concepts.txt, for a description of Unionfs's behavior when mixing read-only
426 +and read-write branches and mounts.
427 +
428 +Syntax:
429 +
430 + dirs=/branch1[=ro|=rw]:/branch2[=ro|=rw]:...:/branchN[=ro|=rw]
431 +
432 +Example:
433 +
434 + dirs=/writable_branch=rw:/read-only_branch=ro
435 +
436 +
437 +BRANCH MANAGEMENT
438 +=================
439 +
440 +Once you mount your union for the first time, using the "dirs=" option, you
441 +can then change the union's overall mode or reconfigure the branches, using
442 +the remount option, as follows.
443 +
444 +To downgrade a union from read-write to read-only:
445 +
446 +# mount -t unionfs -o remount,ro none MOUNTPOINT
447 +
448 +To upgrade a union from read-only to read-write:
449 +
450 +# mount -t unionfs -o remount,rw none MOUNTPOINT
451 +
452 +To delete a branch /foo, regardless where it is in the current union:
453 +
454 +# mount -t unionfs -o remount,del=/foo none MOUNTPOINT
455 +
456 +To insert (add) a branch /foo before /bar:
457 +
458 +# mount -t unionfs -o remount,add=/bar:/foo none MOUNTPOINT
459 +
460 +To insert (add) a branch /foo (with the "rw" mode flag) before /bar:
461 +
462 +# mount -t unionfs -o remount,add=/bar:/foo=rw none MOUNTPOINT
463 +
464 +To insert (add) a branch /foo (in "rw" mode) at the very beginning (i.e., a
465 +new highest-priority branch), you can use the above syntax, or use a short
466 +hand version as follows:
467 +
468 +# mount -t unionfs -o remount,add=/foo none MOUNTPOINT
469 +
470 +To append a branch to the very end (new lowest-priority branch):
471 +
472 +# mount -t unionfs -o remount,add=:/foo none MOUNTPOINT
473 +
474 +To append a branch to the very end (new lowest-priority branch), in
475 +read-only mode:
476 +
477 +# mount -t unionfs -o remount,add=:/foo=ro none MOUNTPOINT
478 +
479 +Finally, to change the mode of one existing branch, say /foo, from read-only
480 +to read-write, and change /bar from read-write to read-only:
481 +
482 +# mount -t unionfs -o remount,mode=/foo=rw,mode=/bar=ro none MOUNTPOINT
483 +
484 +Note: in Unionfs 2.x, you cannot set the leftmost branch to readonly because
485 +then Unionfs won't have any writable place for copyups to take place.
486 +Moreover, the VFS can get confused when it tries to modify something in a
487 +file system mounted read-write, but isn't permitted to write to it.
488 +Instead, you should set the whole union as readonly, as described above.
489 +If, however, you must set the leftmost branch as readonly, perhaps so you
490 +can get a snapshot of it at a point in time, then you should insert a new
491 +writable top-level branch, and mark the one you want as readonly. This can
492 +be accomplished as follows, assuming that /foo is your current leftmost
493 +branch:
494 +
495 +# mount -t tmpfs -o size=NNN /new
496 +# mount -t unionfs -o remount,add=/new,mode=/foo=ro none MOUNTPOINT
497 +<do what you want safely in /foo>
498 +# mount -t unionfs -o remount,del=/new,mode=/foo=rw none MOUNTPOINT
499 +<check if there's anything in /new you want to preserve>
500 +# umount /new
501 +
502 +CACHE CONSISTENCY
503 +=================
504 +
505 +If you modify any file on any of the lower branches directly, while there is
506 +a Unionfs 2.x mounted above any of those branches, you should tell Unionfs
507 +to purge its caches and re-get the objects. To do that, you have to
508 +increment the generation number of the superblock using the following
509 +command:
510 +
511 +# mount -t unionfs -o remount,incgen none MOUNTPOINT
512 +
513 +Note that the older way of incrementing the generation number using an
514 +ioctl, is no longer supported in Unionfs 2.0 and newer. Ioctls in general
515 +are not encouraged. Plus, an ioctl is per-file concept, whereas the
516 +generation number is a per-file-system concept. Worse, such an ioctl
517 +requires an open file, which then has to be invalidated by the very nature
518 +of the generation number increase (read: the old generation increase ioctl
519 +was pretty racy).
520 +
521 +
522 +For more information, see <http://unionfs.filesystems.org/>.
523 diff -Naur linux-2.6.30/fs/ecryptfs/dentry.c linux-2.6.30-magellan/fs/ecryptfs/dentry.c
524 --- linux-2.6.30/fs/ecryptfs/dentry.c 2009-06-10 05:05:27.000000000 +0200
525 +++ linux-2.6.30-magellan/fs/ecryptfs/dentry.c 2009-06-11 18:39:19.000000000 +0200
526 @@ -62,7 +62,7 @@
527 struct inode *lower_inode =
528 ecryptfs_inode_to_lower(dentry->d_inode);
529
530 - fsstack_copy_attr_all(dentry->d_inode, lower_inode, NULL);
531 + fsstack_copy_attr_all(dentry->d_inode, lower_inode);
532 }
533 out:
534 return rc;
535 diff -Naur linux-2.6.30/fs/ecryptfs/inode.c linux-2.6.30-magellan/fs/ecryptfs/inode.c
536 --- linux-2.6.30/fs/ecryptfs/inode.c 2009-06-10 05:05:27.000000000 +0200
537 +++ linux-2.6.30-magellan/fs/ecryptfs/inode.c 2009-06-11 18:39:19.000000000 +0200
538 @@ -624,9 +624,9 @@
539 lower_new_dir_dentry->d_inode, lower_new_dentry);
540 if (rc)
541 goto out_lock;
542 - fsstack_copy_attr_all(new_dir, lower_new_dir_dentry->d_inode, NULL);
543 + fsstack_copy_attr_all(new_dir, lower_new_dir_dentry->d_inode);
544 if (new_dir != old_dir)
545 - fsstack_copy_attr_all(old_dir, lower_old_dir_dentry->d_inode, NULL);
546 + fsstack_copy_attr_all(old_dir, lower_old_dir_dentry->d_inode);
547 out_lock:
548 unlock_rename(lower_old_dir_dentry, lower_new_dir_dentry);
549 dput(lower_new_dentry->d_parent);
550 @@ -965,7 +965,7 @@
551 rc = notify_change(lower_dentry, ia);
552 mutex_unlock(&lower_dentry->d_inode->i_mutex);
553 out:
554 - fsstack_copy_attr_all(inode, lower_inode, NULL);
555 + fsstack_copy_attr_all(inode, lower_inode);
556 return rc;
557 }
558
559 diff -Naur linux-2.6.30/fs/ecryptfs/main.c linux-2.6.30-magellan/fs/ecryptfs/main.c
560 --- linux-2.6.30/fs/ecryptfs/main.c 2009-06-10 05:05:27.000000000 +0200
561 +++ linux-2.6.30-magellan/fs/ecryptfs/main.c 2009-06-11 18:42:20.000000000 +0200
562 @@ -190,7 +190,7 @@
563 init_special_inode(inode, lower_inode->i_mode,
564 lower_inode->i_rdev);
565 dentry->d_op = &ecryptfs_dops;
566 - fsstack_copy_attr_all(inode, lower_inode, NULL);
567 + fsstack_copy_attr_all(inode, lower_inode);
568 /* This size will be overwritten for real files w/ headers and
569 * other metadata */
570 fsstack_copy_inode_size(inode, lower_inode);
571 diff -Naur linux-2.6.30/fs/Kconfig linux-2.6.30-magellan/fs/Kconfig
572 --- linux-2.6.30/fs/Kconfig 2009-06-10 05:05:27.000000000 +0200
573 +++ linux-2.6.30-magellan/fs/Kconfig 2009-06-11 18:39:19.000000000 +0200
574 @@ -139,6 +139,36 @@
575
576 endmenu
577
578 +
579 +menu "Layered filesystems"
580 +
581 +config UNION_FS
582 + tristate "Union file system (EXPERIMENTAL)"
583 + depends on EXPERIMENTAL
584 + help
585 + Unionfs is a stackable unification file system, which appears to
586 + merge the contents of several directories (branches), while keeping
587 + their physical content separate.
588 +
589 + See <http://unionfs.filesystems.org/> for details.
590 +
591 +config UNION_FS_XATTR
592 + bool "Unionfs extended attributes"
593 + depends on UNION_FS
594 + help
595 + Extended attributes are name:value pairs associated with inodes by
596 + the kernel or by users (see the attr(5) manual page).
597 +
598 + If unsure, say N.
599 +
600 +config UNION_FS_DEBUG
601 + bool "Debug Unionfs"
602 + depends on UNION_FS
603 + help
604 + If you say Y here, you can turn on debugging output from Unionfs.
605 +
606 +endmenu
607 +
608 menuconfig MISC_FILESYSTEMS
609 bool "Miscellaneous filesystems"
610 default y
611 diff -Naur linux-2.6.30/fs/Makefile linux-2.6.30-magellan/fs/Makefile
612 --- linux-2.6.30/fs/Makefile 2009-06-10 05:05:27.000000000 +0200
613 +++ linux-2.6.30-magellan/fs/Makefile 2009-06-11 18:39:19.000000000 +0200
614 @@ -84,6 +84,7 @@
615 obj-$(CONFIG_HFSPLUS_FS) += hfsplus/ # Before hfs to find wrapped HFS+
616 obj-$(CONFIG_HFS_FS) += hfs/
617 obj-$(CONFIG_ECRYPT_FS) += ecryptfs/
618 +obj-$(CONFIG_UNION_FS) += unionfs/
619 obj-$(CONFIG_VXFS_FS) += freevxfs/
620 obj-$(CONFIG_NFS_FS) += nfs/
621 obj-$(CONFIG_EXPORTFS) += exportfs/
622 diff -Naur linux-2.6.30/fs/namei.c linux-2.6.30-magellan/fs/namei.c
623 --- linux-2.6.30/fs/namei.c 2009-06-10 05:05:27.000000000 +0200
624 +++ linux-2.6.30-magellan/fs/namei.c 2009-06-11 18:39:19.000000000 +0200
625 @@ -375,6 +375,7 @@
626 else
627 fput(nd->intent.open.file);
628 }
629 +EXPORT_SYMBOL_GPL(release_open_intent);
630
631 static inline struct dentry *
632 do_revalidate(struct dentry *dentry, struct nameidata *nd)
633 diff -Naur linux-2.6.30/fs/splice.c linux-2.6.30-magellan/fs/splice.c
634 --- linux-2.6.30/fs/splice.c 2009-06-10 05:05:27.000000000 +0200
635 +++ linux-2.6.30-magellan/fs/splice.c 2009-06-11 18:39:19.000000000 +0200
636 @@ -905,8 +905,8 @@
637 /*
638 * Attempt to initiate a splice from pipe to file.
639 */
640 -static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
641 - loff_t *ppos, size_t len, unsigned int flags)
642 +long vfs_splice_from(struct pipe_inode_info *pipe, struct file *out,
643 + loff_t *ppos, size_t len, unsigned int flags)
644 {
645 int ret;
646
647 @@ -925,13 +925,14 @@
648
649 return out->f_op->splice_write(pipe, out, ppos, len, flags);
650 }
651 +EXPORT_SYMBOL_GPL(vfs_splice_from);
652
653 /*
654 * Attempt to initiate a splice from a file to a pipe.
655 */
656 -static long do_splice_to(struct file *in, loff_t *ppos,
657 - struct pipe_inode_info *pipe, size_t len,
658 - unsigned int flags)
659 +long vfs_splice_to(struct file *in, loff_t *ppos,
660 + struct pipe_inode_info *pipe, size_t len,
661 + unsigned int flags)
662 {
663 int ret;
664
665 @@ -947,6 +948,7 @@
666
667 return in->f_op->splice_read(in, ppos, pipe, len, flags);
668 }
669 +EXPORT_SYMBOL_GPL(vfs_splice_to);
670
671 /**
672 * splice_direct_to_actor - splices data directly between two non-pipes
673 @@ -1016,7 +1018,7 @@
674 size_t read_len;
675 loff_t pos = sd->pos, prev_pos = pos;
676
677 - ret = do_splice_to(in, &pos, pipe, len, flags);
678 + ret = vfs_splice_to(in, &pos, pipe, len, flags);
679 if (unlikely(ret <= 0))
680 goto out_release;
681
682 @@ -1075,7 +1077,7 @@
683 {
684 struct file *file = sd->u.file;
685
686 - return do_splice_from(pipe, file, &sd->pos, sd->total_len, sd->flags);
687 + return vfs_splice_from(pipe, file, &sd->pos, sd->total_len, sd->flags);
688 }
689
690 /**
691 @@ -1149,7 +1151,7 @@
692 } else
693 off = &out->f_pos;
694
695 - ret = do_splice_from(pipe, out, off, len, flags);
696 + ret = vfs_splice_from(pipe, out, off, len, flags);
697
698 if (off_out && copy_to_user(off_out, off, sizeof(loff_t)))
699 ret = -EFAULT;
700 @@ -1170,7 +1172,7 @@
701 } else
702 off = &in->f_pos;
703
704 - ret = do_splice_to(in, off, pipe, len, flags);
705 + ret = vfs_splice_to(in, off, pipe, len, flags);
706
707 if (off_in && copy_to_user(off_in, off, sizeof(loff_t)))
708 ret = -EFAULT;
709 diff -Naur linux-2.6.30/fs/stack.c linux-2.6.30-magellan/fs/stack.c
710 --- linux-2.6.30/fs/stack.c 2009-06-10 05:05:27.000000000 +0200
711 +++ linux-2.6.30-magellan/fs/stack.c 2009-06-11 18:39:19.000000000 +0200
712 @@ -1,24 +1,82 @@
713 +/*
714 + * Copyright (c) 2006-2009 Erez Zadok
715 + * Copyright (c) 2006-2007 Josef 'Jeff' Sipek
716 + * Copyright (c) 2006-2009 Stony Brook University
717 + * Copyright (c) 2006-2009 The Research Foundation of SUNY
718 + *
719 + * This program is free software; you can redistribute it and/or modify
720 + * it under the terms of the GNU General Public License version 2 as
721 + * published by the Free Software Foundation.
722 + */
723 +
724 #include <linux/module.h>
725 #include <linux/fs.h>
726 #include <linux/fs_stack.h>
727
728 -/* does _NOT_ require i_mutex to be held.
729 +/*
730 + * does _NOT_ require i_mutex to be held.
731 *
732 * This function cannot be inlined since i_size_{read,write} is rather
733 * heavy-weight on 32-bit systems
734 */
735 -void fsstack_copy_inode_size(struct inode *dst, const struct inode *src)
736 +void fsstack_copy_inode_size(struct inode *dst, struct inode *src)
737 {
738 - i_size_write(dst, i_size_read((struct inode *)src));
739 - dst->i_blocks = src->i_blocks;
740 + loff_t i_size;
741 + blkcnt_t i_blocks;
742 +
743 + /*
744 + * i_size_read() includes its own seqlocking and protection from
745 + * preemption (see include/linux/fs.h): we need nothing extra for
746 + * that here, and prefer to avoid nesting locks than attempt to
747 + * keep i_size and i_blocks in synch together.
748 + */
749 + i_size = i_size_read(src);
750 +
751 + /*
752 + * But if CONFIG_LSF (on 32-bit), we ought to make an effort to keep
753 + * the two halves of i_blocks in synch despite SMP or PREEMPT - though
754 + * stat's generic_fillattr() doesn't bother, and we won't be applying
755 + * quotas (where i_blocks does become important) at the upper level.
756 + *
757 + * We don't actually know what locking is used at the lower level; but
758 + * if it's a filesystem that supports quotas, it will be using i_lock
759 + * as in inode_add_bytes(). tmpfs uses other locking, and its 32-bit
760 + * is (just) able to exceed 2TB i_size with the aid of holes; but its
761 + * i_blocks cannot carry into the upper long without almost 2TB swap -
762 + * let's ignore that case.
763 + */
764 + if (sizeof(i_blocks) > sizeof(long))
765 + spin_lock(&src->i_lock);
766 + i_blocks = src->i_blocks;
767 + if (sizeof(i_blocks) > sizeof(long))
768 + spin_unlock(&src->i_lock);
769 +
770 + /*
771 + * If CONFIG_SMP on 32-bit, it's vital for fsstack_copy_inode_size()
772 + * to hold some lock around i_size_write(), otherwise i_size_read()
773 + * may spin forever (see include/linux/fs.h). We don't necessarily
774 + * hold i_mutex when this is called, so take i_lock for that case.
775 + *
776 + * And if CONFIG_LSF (on 32-bit), continue our effort to keep the
777 + * two halves of i_blocks in synch despite SMP or PREEMPT: use i_lock
778 + * for that case too, and do both at once by combining the tests.
779 + *
780 + * There is none of this locking overhead in the 64-bit case.
781 + */
782 + if (sizeof(i_size) > sizeof(long) || sizeof(i_blocks) > sizeof(long))
783 + spin_lock(&dst->i_lock);
784 + i_size_write(dst, i_size);
785 + dst->i_blocks = i_blocks;
786 + if (sizeof(i_size) > sizeof(long) || sizeof(i_blocks) > sizeof(long))
787 + spin_unlock(&dst->i_lock);
788 }
789 EXPORT_SYMBOL_GPL(fsstack_copy_inode_size);
790
791 -/* copy all attributes; get_nlinks is optional way to override the i_nlink
792 +/*
793 + * copy all attributes; get_nlinks is optional way to override the i_nlink
794 * copying
795 */
796 -void fsstack_copy_attr_all(struct inode *dest, const struct inode *src,
797 - int (*get_nlinks)(struct inode *))
798 +void fsstack_copy_attr_all(struct inode *dest, const struct inode *src)
799 {
800 dest->i_mode = src->i_mode;
801 dest->i_uid = src->i_uid;
802 @@ -29,14 +87,6 @@
803 dest->i_ctime = src->i_ctime;
804 dest->i_blkbits = src->i_blkbits;
805 dest->i_flags = src->i_flags;
806 -
807 - /*
808 - * Update the nlinks AFTER updating the above fields, because the
809 - * get_links callback may depend on them.
810 - */
811 - if (!get_nlinks)
812 - dest->i_nlink = src->i_nlink;
813 - else
814 - dest->i_nlink = (*get_nlinks)(dest);
815 + dest->i_nlink = src->i_nlink;
816 }
817 EXPORT_SYMBOL_GPL(fsstack_copy_attr_all);
818 diff -Naur linux-2.6.30/fs/unionfs/commonfops.c linux-2.6.30-magellan/fs/unionfs/commonfops.c
819 --- linux-2.6.30/fs/unionfs/commonfops.c 1970-01-01 01:00:00.000000000 +0100
820 +++ linux-2.6.30-magellan/fs/unionfs/commonfops.c 2009-06-11 18:39:19.000000000 +0200
821 @@ -0,0 +1,880 @@
822 +/*
823 + * Copyright (c) 2003-2009 Erez Zadok
824 + * Copyright (c) 2003-2006 Charles P. Wright
825 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
826 + * Copyright (c) 2005-2006 Junjiro Okajima
827 + * Copyright (c) 2005 Arun M. Krishnakumar
828 + * Copyright (c) 2004-2006 David P. Quigley
829 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
830 + * Copyright (c) 2003 Puja Gupta
831 + * Copyright (c) 2003 Harikesavan Krishnan
832 + * Copyright (c) 2003-2009 Stony Brook University
833 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
834 + *
835 + * This program is free software; you can redistribute it and/or modify
836 + * it under the terms of the GNU General Public License version 2 as
837 + * published by the Free Software Foundation.
838 + */
839 +
840 +#include "union.h"
841 +
842 +/*
843 + * 1) Copyup the file
844 + * 2) Rename the file to '.unionfs<original inode#><counter>' - obviously
845 + * stolen from NFS's silly rename
846 + */
847 +static int copyup_deleted_file(struct file *file, struct dentry *dentry,
848 + struct dentry *parent, int bstart, int bindex)
849 +{
850 + static unsigned int counter;
851 + const int i_inosize = sizeof(dentry->d_inode->i_ino) * 2;
852 + const int countersize = sizeof(counter) * 2;
853 + const int nlen = sizeof(".unionfs") + i_inosize + countersize - 1;
854 + char name[nlen + 1];
855 + int err;
856 + struct dentry *tmp_dentry = NULL;
857 + struct dentry *lower_dentry;
858 + struct dentry *lower_dir_dentry = NULL;
859 +
860 + lower_dentry = unionfs_lower_dentry_idx(dentry, bstart);
861 +
862 + sprintf(name, ".unionfs%*.*lx",
863 + i_inosize, i_inosize, lower_dentry->d_inode->i_ino);
864 +
865 + /*
866 + * Loop, looking for an unused temp name to copyup to.
867 + *
868 + * It's somewhat silly that we look for a free temp tmp name in the
869 + * source branch (bstart) instead of the dest branch (bindex), where
870 + * the final name will be created. We _will_ catch it if somehow
871 + * the name exists in the dest branch, but it'd be nice to catch it
872 + * sooner than later.
873 + */
874 +retry:
875 + tmp_dentry = NULL;
876 + do {
877 + char *suffix = name + nlen - countersize;
878 +
879 + dput(tmp_dentry);
880 + counter++;
881 + sprintf(suffix, "%*.*x", countersize, countersize, counter);
882 +
883 + pr_debug("unionfs: trying to rename %s to %s\n",
884 + dentry->d_name.name, name);
885 +
886 + tmp_dentry = lookup_one_len(name, lower_dentry->d_parent,
887 + nlen);
888 + if (IS_ERR(tmp_dentry)) {
889 + err = PTR_ERR(tmp_dentry);
890 + goto out;
891 + }
892 + } while (tmp_dentry->d_inode != NULL); /* need negative dentry */
893 + dput(tmp_dentry);
894 +
895 + err = copyup_named_file(parent->d_inode, file, name, bstart, bindex,
896 + i_size_read(file->f_path.dentry->d_inode));
897 + if (err) {
898 + if (unlikely(err == -EEXIST))
899 + goto retry;
900 + goto out;
901 + }
902 +
903 + /* bring it to the same state as an unlinked file */
904 + lower_dentry = unionfs_lower_dentry_idx(dentry, dbstart(dentry));
905 + if (!unionfs_lower_inode_idx(dentry->d_inode, bindex)) {
906 + atomic_inc(&lower_dentry->d_inode->i_count);
907 + unionfs_set_lower_inode_idx(dentry->d_inode, bindex,
908 + lower_dentry->d_inode);
909 + }
910 + lower_dir_dentry = lock_parent(lower_dentry);
911 + err = vfs_unlink(lower_dir_dentry->d_inode, lower_dentry);
912 + unlock_dir(lower_dir_dentry);
913 +
914 +out:
915 + if (!err)
916 + unionfs_check_dentry(dentry);
917 + return err;
918 +}
919 +
920 +/*
921 + * put all references held by upper struct file and free lower file pointer
922 + * array
923 + */
924 +static void cleanup_file(struct file *file)
925 +{
926 + int bindex, bstart, bend;
927 + struct file **lower_files;
928 + struct file *lower_file;
929 + struct super_block *sb = file->f_path.dentry->d_sb;
930 +
931 + lower_files = UNIONFS_F(file)->lower_files;
932 + bstart = fbstart(file);
933 + bend = fbend(file);
934 +
935 + for (bindex = bstart; bindex <= bend; bindex++) {
936 + int i; /* holds (possibly) updated branch index */
937 + int old_bid;
938 +
939 + lower_file = unionfs_lower_file_idx(file, bindex);
940 + if (!lower_file)
941 + continue;
942 +
943 + /*
944 + * Find new index of matching branch with an open
945 + * file, since branches could have been added or
946 + * deleted causing the one with open files to shift.
947 + */
948 + old_bid = UNIONFS_F(file)->saved_branch_ids[bindex];
949 + i = branch_id_to_idx(sb, old_bid);
950 + if (unlikely(i < 0)) {
951 + printk(KERN_ERR "unionfs: no superblock for "
952 + "file %p\n", file);
953 + continue;
954 + }
955 +
956 + /* decrement count of open files */
957 + branchput(sb, i);
958 + /*
959 + * fput will perform an mntput for us on the correct branch.
960 + * Although we're using the file's old branch configuration,
961 + * bindex, which is the old index, correctly points to the
962 + * right branch in the file's branch list. In other words,
963 + * we're going to mntput the correct branch even if branches
964 + * have been added/removed.
965 + */
966 + fput(lower_file);
967 + UNIONFS_F(file)->lower_files[bindex] = NULL;
968 + UNIONFS_F(file)->saved_branch_ids[bindex] = -1;
969 + }
970 +
971 + UNIONFS_F(file)->lower_files = NULL;
972 + kfree(lower_files);
973 + kfree(UNIONFS_F(file)->saved_branch_ids);
974 + /* set to NULL because caller needs to know if to kfree on error */
975 + UNIONFS_F(file)->saved_branch_ids = NULL;
976 +}
977 +
978 +/* open all lower files for a given file */
979 +static int open_all_files(struct file *file)
980 +{
981 + int bindex, bstart, bend, err = 0;
982 + struct file *lower_file;
983 + struct dentry *lower_dentry;
984 + struct dentry *dentry = file->f_path.dentry;
985 + struct super_block *sb = dentry->d_sb;
986 +
987 + bstart = dbstart(dentry);
988 + bend = dbend(dentry);
989 +
990 + for (bindex = bstart; bindex <= bend; bindex++) {
991 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
992 + if (!lower_dentry)
993 + continue;
994 +
995 + dget(lower_dentry);
996 + unionfs_mntget(dentry, bindex);
997 + branchget(sb, bindex);
998 +
999 + lower_file =
1000 + dentry_open(lower_dentry,
1001 + unionfs_lower_mnt_idx(dentry, bindex),
1002 + file->f_flags, current_cred());
1003 + if (IS_ERR(lower_file)) {
1004 + branchput(sb, bindex);
1005 + err = PTR_ERR(lower_file);
1006 + goto out;
1007 + } else {
1008 + unionfs_set_lower_file_idx(file, bindex, lower_file);
1009 + }
1010 + }
1011 +out:
1012 + return err;
1013 +}
1014 +
1015 +/* open the highest priority file for a given upper file */
1016 +static int open_highest_file(struct file *file, bool willwrite)
1017 +{
1018 + int bindex, bstart, bend, err = 0;
1019 + struct file *lower_file;
1020 + struct dentry *lower_dentry;
1021 + struct dentry *dentry = file->f_path.dentry;
1022 + struct dentry *parent = dget_parent(dentry);
1023 + struct inode *parent_inode = parent->d_inode;
1024 + struct super_block *sb = dentry->d_sb;
1025 +
1026 + bstart = dbstart(dentry);
1027 + bend = dbend(dentry);
1028 +
1029 + lower_dentry = unionfs_lower_dentry(dentry);
1030 + if (willwrite && IS_WRITE_FLAG(file->f_flags) && is_robranch(dentry)) {
1031 + for (bindex = bstart - 1; bindex >= 0; bindex--) {
1032 + err = copyup_file(parent_inode, file, bstart, bindex,
1033 + i_size_read(dentry->d_inode));
1034 + if (!err)
1035 + break;
1036 + }
1037 + atomic_set(&UNIONFS_F(file)->generation,
1038 + atomic_read(&UNIONFS_I(dentry->d_inode)->
1039 + generation));
1040 + goto out;
1041 + }
1042 +
1043 + dget(lower_dentry);
1044 + unionfs_mntget(dentry, bstart);
1045 + lower_file = dentry_open(lower_dentry,
1046 + unionfs_lower_mnt_idx(dentry, bstart),
1047 + file->f_flags, current_cred());
1048 + if (IS_ERR(lower_file)) {
1049 + err = PTR_ERR(lower_file);
1050 + goto out;
1051 + }
1052 + branchget(sb, bstart);
1053 + unionfs_set_lower_file(file, lower_file);
1054 + /* Fix up the position. */
1055 + lower_file->f_pos = file->f_pos;
1056 +
1057 + memcpy(&lower_file->f_ra, &file->f_ra, sizeof(struct file_ra_state));
1058 +out:
1059 + dput(parent);
1060 + return err;
1061 +}
1062 +
1063 +/* perform a delayed copyup of a read-write file on a read-only branch */
1064 +static int do_delayed_copyup(struct file *file, struct dentry *parent)
1065 +{
1066 + int bindex, bstart, bend, err = 0;
1067 + struct dentry *dentry = file->f_path.dentry;
1068 + struct inode *parent_inode = parent->d_inode;
1069 +
1070 + bstart = fbstart(file);
1071 + bend = fbend(file);
1072 +
1073 + BUG_ON(!S_ISREG(dentry->d_inode->i_mode));
1074 +
1075 + unionfs_check_file(file);
1076 + for (bindex = bstart - 1; bindex >= 0; bindex--) {
1077 + if (!d_deleted(dentry))
1078 + err = copyup_file(parent_inode, file, bstart,
1079 + bindex,
1080 + i_size_read(dentry->d_inode));
1081 + else
1082 + err = copyup_deleted_file(file, dentry, parent,
1083 + bstart, bindex);
1084 + /* if succeeded, set lower open-file flags and break */
1085 + if (!err) {
1086 + struct file *lower_file;
1087 + lower_file = unionfs_lower_file_idx(file, bindex);
1088 + lower_file->f_flags = file->f_flags;
1089 + break;
1090 + }
1091 + }
1092 + if (err || (bstart <= fbstart(file)))
1093 + goto out;
1094 + bend = fbend(file);
1095 + for (bindex = bstart; bindex <= bend; bindex++) {
1096 + if (unionfs_lower_file_idx(file, bindex)) {
1097 + branchput(dentry->d_sb, bindex);
1098 + fput(unionfs_lower_file_idx(file, bindex));
1099 + unionfs_set_lower_file_idx(file, bindex, NULL);
1100 + }
1101 + }
1102 + path_put_lowers(dentry, bstart, bend, false);
1103 + iput_lowers(dentry->d_inode, bstart, bend, false);
1104 + /* for reg file, we only open it "once" */
1105 + fbend(file) = fbstart(file);
1106 + dbend(dentry) = dbstart(dentry);
1107 + ibend(dentry->d_inode) = ibstart(dentry->d_inode);
1108 +
1109 +out:
1110 + unionfs_check_file(file);
1111 + return err;
1112 +}
1113 +
1114 +/*
1115 + * Helper function for unionfs_file_revalidate/locked.
1116 + * Expects dentry/parent to be locked already, and revalidated.
1117 + */
1118 +static int __unionfs_file_revalidate(struct file *file, struct dentry *dentry,
1119 + struct dentry *parent,
1120 + struct super_block *sb, int sbgen,
1121 + int dgen, bool willwrite)
1122 +{
1123 + int fgen;
1124 + int bstart, bend, orig_brid;
1125 + int size;
1126 + int err = 0;
1127 +
1128 + fgen = atomic_read(&UNIONFS_F(file)->generation);
1129 +
1130 + /*
1131 + * There are two cases we are interested in. The first is if the
1132 + * generation is lower than the super-block. The second is if
1133 + * someone has copied up this file from underneath us, we also need
1134 + * to refresh things.
1135 + */
1136 + if (d_deleted(dentry) ||
1137 + (sbgen <= fgen &&
1138 + dbstart(dentry) == fbstart(file) &&
1139 + unionfs_lower_file(file)))
1140 + goto out_may_copyup;
1141 +
1142 + /* save orig branch ID */
1143 + orig_brid = UNIONFS_F(file)->saved_branch_ids[fbstart(file)];
1144 +
1145 + /* First we throw out the existing files. */
1146 + cleanup_file(file);
1147 +
1148 + /* Now we reopen the file(s) as in unionfs_open. */
1149 + bstart = fbstart(file) = dbstart(dentry);
1150 + bend = fbend(file) = dbend(dentry);
1151 +
1152 + size = sizeof(struct file *) * sbmax(sb);
1153 + UNIONFS_F(file)->lower_files = kzalloc(size, GFP_KERNEL);
1154 + if (unlikely(!UNIONFS_F(file)->lower_files)) {
1155 + err = -ENOMEM;
1156 + goto out;
1157 + }
1158 + size = sizeof(int) * sbmax(sb);
1159 + UNIONFS_F(file)->saved_branch_ids = kzalloc(size, GFP_KERNEL);
1160 + if (unlikely(!UNIONFS_F(file)->saved_branch_ids)) {
1161 + err = -ENOMEM;
1162 + goto out;
1163 + }
1164 +
1165 + if (S_ISDIR(dentry->d_inode->i_mode)) {
1166 + /* We need to open all the files. */
1167 + err = open_all_files(file);
1168 + if (err)
1169 + goto out;
1170 + } else {
1171 + int new_brid;
1172 + /* We only open the highest priority branch. */
1173 + err = open_highest_file(file, willwrite);
1174 + if (err)
1175 + goto out;
1176 + new_brid = UNIONFS_F(file)->saved_branch_ids[fbstart(file)];
1177 + if (unlikely(new_brid != orig_brid && sbgen > fgen)) {
1178 + /*
1179 + * If we re-opened the file on a different branch
1180 + * than the original one, and this was due to a new
1181 + * branch inserted, then update the mnt counts of
1182 + * the old and new branches accordingly.
1183 + */
1184 + unionfs_mntget(dentry, bstart);
1185 + unionfs_mntput(sb->s_root,
1186 + branch_id_to_idx(sb, orig_brid));
1187 + }
1188 + /* regular files have only one open lower file */
1189 + fbend(file) = fbstart(file);
1190 + }
1191 + atomic_set(&UNIONFS_F(file)->generation,
1192 + atomic_read(&UNIONFS_I(dentry->d_inode)->generation));
1193 +
1194 +out_may_copyup:
1195 + /* Copyup on the first write to a file on a readonly branch. */
1196 + if (willwrite && IS_WRITE_FLAG(file->f_flags) &&
1197 + !IS_WRITE_FLAG(unionfs_lower_file(file)->f_flags) &&
1198 + is_robranch(dentry)) {
1199 + pr_debug("unionfs: do delay copyup of \"%s\"\n",
1200 + dentry->d_name.name);
1201 + err = do_delayed_copyup(file, parent);
1202 + /* regular files have only one open lower file */
1203 + if (!err && !S_ISDIR(dentry->d_inode->i_mode))
1204 + fbend(file) = fbstart(file);
1205 + }
1206 +
1207 +out:
1208 + if (err) {
1209 + kfree(UNIONFS_F(file)->lower_files);
1210 + kfree(UNIONFS_F(file)->saved_branch_ids);
1211 + }
1212 + return err;
1213 +}
1214 +
1215 +/*
1216 + * Revalidate the struct file
1217 + * @file: file to revalidate
1218 + * @parent: parent dentry (locked by caller)
1219 + * @willwrite: true if caller may cause changes to the file; false otherwise.
1220 + * Caller must lock/unlock dentry's branch configuration.
1221 + */
1222 +int unionfs_file_revalidate(struct file *file, struct dentry *parent,
1223 + bool willwrite)
1224 +{
1225 + struct super_block *sb;
1226 + struct dentry *dentry;
1227 + int sbgen, dgen;
1228 + int err = 0;
1229 +
1230 + dentry = file->f_path.dentry;
1231 + sb = dentry->d_sb;
1232 + verify_locked(dentry);
1233 + verify_locked(parent);
1234 +
1235 + /*
1236 + * First revalidate the dentry inside struct file,
1237 + * but not unhashed dentries.
1238 + */
1239 + if (!d_deleted(dentry) &&
1240 + !__unionfs_d_revalidate(dentry, parent, willwrite)) {
1241 + err = -ESTALE;
1242 + goto out;
1243 + }
1244 +
1245 + sbgen = atomic_read(&UNIONFS_SB(sb)->generation);
1246 + dgen = atomic_read(&UNIONFS_D(dentry)->generation);
1247 +
1248 + if (unlikely(sbgen > dgen)) { /* XXX: should never happen */
1249 + pr_debug("unionfs: failed to revalidate dentry (%s)\n",
1250 + dentry->d_name.name);
1251 + err = -ESTALE;
1252 + goto out;
1253 + }
1254 +
1255 + err = __unionfs_file_revalidate(file, dentry, parent, sb,
1256 + sbgen, dgen, willwrite);
1257 +out:
1258 + return err;
1259 +}
1260 +
1261 +/* unionfs_open helper function: open a directory */
1262 +static int __open_dir(struct inode *inode, struct file *file)
1263 +{
1264 + struct dentry *lower_dentry;
1265 + struct file *lower_file;
1266 + int bindex, bstart, bend;
1267 + struct vfsmount *mnt;
1268 +
1269 + bstart = fbstart(file) = dbstart(file->f_path.dentry);
1270 + bend = fbend(file) = dbend(file->f_path.dentry);
1271 +
1272 + for (bindex = bstart; bindex <= bend; bindex++) {
1273 + lower_dentry =
1274 + unionfs_lower_dentry_idx(file->f_path.dentry, bindex);
1275 + if (!lower_dentry)
1276 + continue;
1277 +
1278 + dget(lower_dentry);
1279 + unionfs_mntget(file->f_path.dentry, bindex);
1280 + mnt = unionfs_lower_mnt_idx(file->f_path.dentry, bindex);
1281 + lower_file = dentry_open(lower_dentry, mnt, file->f_flags,
1282 + current_cred());
1283 + if (IS_ERR(lower_file))
1284 + return PTR_ERR(lower_file);
1285 +
1286 + unionfs_set_lower_file_idx(file, bindex, lower_file);
1287 +
1288 + /*
1289 + * The branchget goes after the open, because otherwise
1290 + * we would miss the reference on release.
1291 + */
1292 + branchget(inode->i_sb, bindex);
1293 + }
1294 +
1295 + return 0;
1296 +}
1297 +
1298 +/* unionfs_open helper function: open a file */
1299 +static int __open_file(struct inode *inode, struct file *file,
1300 + struct dentry *parent)
1301 +{
1302 + struct dentry *lower_dentry;
1303 + struct file *lower_file;
1304 + int lower_flags;
1305 + int bindex, bstart, bend;
1306 +
1307 + lower_dentry = unionfs_lower_dentry(file->f_path.dentry);
1308 + lower_flags = file->f_flags;
1309 +
1310 + bstart = fbstart(file) = dbstart(file->f_path.dentry);
1311 + bend = fbend(file) = dbend(file->f_path.dentry);
1312 +
1313 + /*
1314 + * check for the permission for lower file. If the error is
1315 + * COPYUP_ERR, copyup the file.
1316 + */
1317 + if (lower_dentry->d_inode && is_robranch(file->f_path.dentry)) {
1318 + /*
1319 + * if the open will change the file, copy it up otherwise
1320 + * defer it.
1321 + */
1322 + if (lower_flags & O_TRUNC) {
1323 + int size = 0;
1324 + int err = -EROFS;
1325 +
1326 + /* copyup the file */
1327 + for (bindex = bstart - 1; bindex >= 0; bindex--) {
1328 + err = copyup_file(parent->d_inode, file,
1329 + bstart, bindex, size);
1330 + if (!err)
1331 + break;
1332 + }
1333 + return err;
1334 + } else {
1335 + /*
1336 + * turn off writeable flags, to force delayed copyup
1337 + * by caller.
1338 + */
1339 + lower_flags &= ~(OPEN_WRITE_FLAGS);
1340 + }
1341 + }
1342 +
1343 + dget(lower_dentry);
1344 +
1345 + /*
1346 + * dentry_open will decrement mnt refcnt if err.
1347 + * otherwise fput() will do an mntput() for us upon file close.
1348 + */
1349 + unionfs_mntget(file->f_path.dentry, bstart);
1350 + lower_file =
1351 + dentry_open(lower_dentry,
1352 + unionfs_lower_mnt_idx(file->f_path.dentry, bstart),
1353 + lower_flags, current_cred());
1354 + if (IS_ERR(lower_file))
1355 + return PTR_ERR(lower_file);
1356 +
1357 + unionfs_set_lower_file(file, lower_file);
1358 + branchget(inode->i_sb, bstart);
1359 +
1360 + return 0;
1361 +}
1362 +
1363 +int unionfs_open(struct inode *inode, struct file *file)
1364 +{
1365 + int err = 0;
1366 + struct file *lower_file = NULL;
1367 + struct dentry *dentry = file->f_path.dentry;
1368 + struct dentry *parent;
1369 + int bindex = 0, bstart = 0, bend = 0;
1370 + int size;
1371 + int valid = 0;
1372 +
1373 + unionfs_read_lock(inode->i_sb, UNIONFS_SMUTEX_PARENT);
1374 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
1375 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
1376 +
1377 + /* don't open unhashed/deleted files */
1378 + if (d_deleted(dentry)) {
1379 + err = -ENOENT;
1380 + goto out_nofree;
1381 + }
1382 +
1383 + /* XXX: should I change 'false' below to the 'willwrite' flag? */
1384 + valid = __unionfs_d_revalidate(dentry, parent, false);
1385 + if (unlikely(!valid)) {
1386 + err = -ESTALE;
1387 + goto out_nofree;
1388 + }
1389 +
1390 + file->private_data =
1391 + kzalloc(sizeof(struct unionfs_file_info), GFP_KERNEL);
1392 + if (unlikely(!UNIONFS_F(file))) {
1393 + err = -ENOMEM;
1394 + goto out_nofree;
1395 + }
1396 + fbstart(file) = -1;
1397 + fbend(file) = -1;
1398 + atomic_set(&UNIONFS_F(file)->generation,
1399 + atomic_read(&UNIONFS_I(inode)->generation));
1400 +
1401 + size = sizeof(struct file *) * sbmax(inode->i_sb);
1402 + UNIONFS_F(file)->lower_files = kzalloc(size, GFP_KERNEL);
1403 + if (unlikely(!UNIONFS_F(file)->lower_files)) {
1404 + err = -ENOMEM;
1405 + goto out;
1406 + }
1407 + size = sizeof(int) * sbmax(inode->i_sb);
1408 + UNIONFS_F(file)->saved_branch_ids = kzalloc(size, GFP_KERNEL);
1409 + if (unlikely(!UNIONFS_F(file)->saved_branch_ids)) {
1410 + err = -ENOMEM;
1411 + goto out;
1412 + }
1413 +
1414 + bstart = fbstart(file) = dbstart(dentry);
1415 + bend = fbend(file) = dbend(dentry);
1416 +
1417 + /*
1418 + * open all directories and make the unionfs file struct point to
1419 + * these lower file structs
1420 + */
1421 + if (S_ISDIR(inode->i_mode))
1422 + err = __open_dir(inode, file); /* open a dir */
1423 + else
1424 + err = __open_file(inode, file, parent); /* open a file */
1425 +
1426 + /* freeing the allocated resources, and fput the opened files */
1427 + if (err) {
1428 + for (bindex = bstart; bindex <= bend; bindex++) {
1429 + lower_file = unionfs_lower_file_idx(file, bindex);
1430 + if (!lower_file)
1431 + continue;
1432 +
1433 + branchput(dentry->d_sb, bindex);
1434 + /* fput calls dput for lower_dentry */
1435 + fput(lower_file);
1436 + }
1437 + }
1438 +
1439 +out:
1440 + if (err) {
1441 + kfree(UNIONFS_F(file)->lower_files);
1442 + kfree(UNIONFS_F(file)->saved_branch_ids);
1443 + kfree(UNIONFS_F(file));
1444 + }
1445 +out_nofree:
1446 + if (!err) {
1447 + unionfs_postcopyup_setmnt(dentry);
1448 + unionfs_copy_attr_times(inode);
1449 + unionfs_check_file(file);
1450 + unionfs_check_inode(inode);
1451 + }
1452 + unionfs_unlock_dentry(dentry);
1453 + unionfs_unlock_parent(dentry, parent);
1454 + unionfs_read_unlock(inode->i_sb);
1455 + return err;
1456 +}
1457 +
1458 +/*
1459 + * release all lower object references & free the file info structure
1460 + *
1461 + * No need to grab sb info's rwsem.
1462 + */
1463 +int unionfs_file_release(struct inode *inode, struct file *file)
1464 +{
1465 + struct file *lower_file = NULL;
1466 + struct unionfs_file_info *fileinfo;
1467 + struct unionfs_inode_info *inodeinfo;
1468 + struct super_block *sb = inode->i_sb;
1469 + struct dentry *dentry = file->f_path.dentry;
1470 + struct dentry *parent;
1471 + int bindex, bstart, bend;
1472 + int fgen, err = 0;
1473 +
1474 + unionfs_read_lock(sb, UNIONFS_SMUTEX_PARENT);
1475 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
1476 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
1477 +
1478 + /*
1479 + * We try to revalidate, but the VFS ignores return return values
1480 + * from file->release, so we must always try to succeed here,
1481 + * including to do the kfree and dput below. So if revalidation
1482 + * failed, all we can do is print some message and keep going.
1483 + */
1484 + err = unionfs_file_revalidate(file, parent,
1485 + UNIONFS_F(file)->wrote_to_file);
1486 + if (!err)
1487 + unionfs_check_file(file);
1488 + fileinfo = UNIONFS_F(file);
1489 + BUG_ON(file->f_path.dentry->d_inode != inode);
1490 + inodeinfo = UNIONFS_I(inode);
1491 +
1492 + /* fput all the lower files */
1493 + fgen = atomic_read(&fileinfo->generation);
1494 + bstart = fbstart(file);
1495 + bend = fbend(file);
1496 +
1497 + for (bindex = bstart; bindex <= bend; bindex++) {
1498 + lower_file = unionfs_lower_file_idx(file, bindex);
1499 +
1500 + if (lower_file) {
1501 + unionfs_set_lower_file_idx(file, bindex, NULL);
1502 + fput(lower_file);
1503 + branchput(sb, bindex);
1504 + }
1505 +
1506 + /* if there are no more refs to the dentry, dput it */
1507 + if (d_deleted(dentry)) {
1508 + dput(unionfs_lower_dentry_idx(dentry, bindex));
1509 + unionfs_set_lower_dentry_idx(dentry, bindex, NULL);
1510 + }
1511 + }
1512 +
1513 + kfree(fileinfo->lower_files);
1514 + kfree(fileinfo->saved_branch_ids);
1515 +
1516 + if (fileinfo->rdstate) {
1517 + fileinfo->rdstate->access = jiffies;
1518 + spin_lock(&inodeinfo->rdlock);
1519 + inodeinfo->rdcount++;
1520 + list_add_tail(&fileinfo->rdstate->cache,
1521 + &inodeinfo->readdircache);
1522 + mark_inode_dirty(inode);
1523 + spin_unlock(&inodeinfo->rdlock);
1524 + fileinfo->rdstate = NULL;
1525 + }
1526 + kfree(fileinfo);
1527 +
1528 + unionfs_unlock_dentry(dentry);
1529 + unionfs_unlock_parent(dentry, parent);
1530 + unionfs_read_unlock(sb);
1531 + return err;
1532 +}
1533 +
1534 +/* pass the ioctl to the lower fs */
1535 +static long do_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1536 +{
1537 + struct file *lower_file;
1538 + int err;
1539 +
1540 + lower_file = unionfs_lower_file(file);
1541 +
1542 + err = -ENOTTY;
1543 + if (!lower_file || !lower_file->f_op)
1544 + goto out;
1545 + if (lower_file->f_op->unlocked_ioctl) {
1546 + err = lower_file->f_op->unlocked_ioctl(lower_file, cmd, arg);
1547 + } else if (lower_file->f_op->ioctl) {
1548 + lock_kernel();
1549 + err = lower_file->f_op->ioctl(
1550 + lower_file->f_path.dentry->d_inode,
1551 + lower_file, cmd, arg);
1552 + unlock_kernel();
1553 + }
1554 +
1555 +out:
1556 + return err;
1557 +}
1558 +
1559 +/*
1560 + * return to user-space the branch indices containing the file in question
1561 + *
1562 + * We use fd_set and therefore we are limited to the number of the branches
1563 + * to FD_SETSIZE, which is currently 1024 - plenty for most people
1564 + */
1565 +static int unionfs_ioctl_queryfile(struct file *file, struct dentry *parent,
1566 + unsigned int cmd, unsigned long arg)
1567 +{
1568 + int err = 0;
1569 + fd_set branchlist;
1570 + int bstart = 0, bend = 0, bindex = 0;
1571 + int orig_bstart, orig_bend;
1572 + struct dentry *dentry, *lower_dentry;
1573 + struct vfsmount *mnt;
1574 +
1575 + dentry = file->f_path.dentry;
1576 + orig_bstart = dbstart(dentry);
1577 + orig_bend = dbend(dentry);
1578 + err = unionfs_partial_lookup(dentry, parent);
1579 + if (err)
1580 + goto out;
1581 + bstart = dbstart(dentry);
1582 + bend = dbend(dentry);
1583 +
1584 + FD_ZERO(&branchlist);
1585 +
1586 + for (bindex = bstart; bindex <= bend; bindex++) {
1587 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
1588 + if (!lower_dentry)
1589 + continue;
1590 + if (likely(lower_dentry->d_inode))
1591 + FD_SET(bindex, &branchlist);
1592 + /* purge any lower objects after partial_lookup */
1593 + if (bindex < orig_bstart || bindex > orig_bend) {
1594 + dput(lower_dentry);
1595 + unionfs_set_lower_dentry_idx(dentry, bindex, NULL);
1596 + iput(unionfs_lower_inode_idx(dentry->d_inode, bindex));
1597 + unionfs_set_lower_inode_idx(dentry->d_inode, bindex,
1598 + NULL);
1599 + mnt = unionfs_lower_mnt_idx(dentry, bindex);
1600 + if (!mnt)
1601 + continue;
1602 + unionfs_mntput(dentry, bindex);
1603 + unionfs_set_lower_mnt_idx(dentry, bindex, NULL);
1604 + }
1605 + }
1606 + /* restore original dentry's offsets */
1607 + dbstart(dentry) = orig_bstart;
1608 + dbend(dentry) = orig_bend;
1609 + ibstart(dentry->d_inode) = orig_bstart;
1610 + ibend(dentry->d_inode) = orig_bend;
1611 +
1612 + err = copy_to_user((void __user *)arg, &branchlist, sizeof(fd_set));
1613 + if (unlikely(err))
1614 + err = -EFAULT;
1615 +
1616 +out:
1617 + return err < 0 ? err : bend;
1618 +}
1619 +
1620 +long unionfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1621 +{
1622 + long err;
1623 + struct dentry *dentry = file->f_path.dentry;
1624 + struct dentry *parent;
1625 +
1626 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
1627 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
1628 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
1629 +
1630 + err = unionfs_file_revalidate(file, parent, true);
1631 + if (unlikely(err))
1632 + goto out;
1633 +
1634 + /* check if asked for local commands */
1635 + switch (cmd) {
1636 + case UNIONFS_IOCTL_INCGEN:
1637 + /* Increment the superblock generation count */
1638 + pr_info("unionfs: incgen ioctl deprecated; "
1639 + "use \"-o remount,incgen\"\n");
1640 + err = -ENOSYS;
1641 + break;
1642 +
1643 + case UNIONFS_IOCTL_QUERYFILE:
1644 + /* Return list of branches containing the given file */
1645 + err = unionfs_ioctl_queryfile(file, parent, cmd, arg);
1646 + break;
1647 +
1648 + default:
1649 + /* pass the ioctl down */
1650 + err = do_ioctl(file, cmd, arg);
1651 + break;
1652 + }
1653 +
1654 +out:
1655 + unionfs_check_file(file);
1656 + unionfs_unlock_dentry(dentry);
1657 + unionfs_unlock_parent(dentry, parent);
1658 + unionfs_read_unlock(dentry->d_sb);
1659 + return err;
1660 +}
1661 +
1662 +int unionfs_flush(struct file *file, fl_owner_t id)
1663 +{
1664 + int err = 0;
1665 + struct file *lower_file = NULL;
1666 + struct dentry *dentry = file->f_path.dentry;
1667 + struct dentry *parent;
1668 + int bindex, bstart, bend;
1669 +
1670 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
1671 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
1672 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
1673 +
1674 + err = unionfs_file_revalidate(file, parent,
1675 + UNIONFS_F(file)->wrote_to_file);
1676 + if (unlikely(err))
1677 + goto out;
1678 + unionfs_check_file(file);
1679 +
1680 + bstart = fbstart(file);
1681 + bend = fbend(file);
1682 + for (bindex = bstart; bindex <= bend; bindex++) {
1683 + lower_file = unionfs_lower_file_idx(file, bindex);
1684 +
1685 + if (lower_file && lower_file->f_op &&
1686 + lower_file->f_op->flush) {
1687 + err = lower_file->f_op->flush(lower_file, id);
1688 + if (err)
1689 + goto out;
1690 + }
1691 +
1692 + }
1693 +
1694 +out:
1695 + if (!err)
1696 + unionfs_check_file(file);
1697 + unionfs_unlock_dentry(dentry);
1698 + unionfs_unlock_parent(dentry, parent);
1699 + unionfs_read_unlock(dentry->d_sb);
1700 + return err;
1701 +}
1702 diff -Naur linux-2.6.30/fs/unionfs/copyup.c linux-2.6.30-magellan/fs/unionfs/copyup.c
1703 --- linux-2.6.30/fs/unionfs/copyup.c 1970-01-01 01:00:00.000000000 +0100
1704 +++ linux-2.6.30-magellan/fs/unionfs/copyup.c 2009-06-11 18:39:19.000000000 +0200
1705 @@ -0,0 +1,897 @@
1706 +/*
1707 + * Copyright (c) 2003-2009 Erez Zadok
1708 + * Copyright (c) 2003-2006 Charles P. Wright
1709 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
1710 + * Copyright (c) 2005-2006 Junjiro Okajima
1711 + * Copyright (c) 2005 Arun M. Krishnakumar
1712 + * Copyright (c) 2004-2006 David P. Quigley
1713 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
1714 + * Copyright (c) 2003 Puja Gupta
1715 + * Copyright (c) 2003 Harikesavan Krishnan
1716 + * Copyright (c) 2003-2009 Stony Brook University
1717 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
1718 + *
1719 + * This program is free software; you can redistribute it and/or modify
1720 + * it under the terms of the GNU General Public License version 2 as
1721 + * published by the Free Software Foundation.
1722 + */
1723 +
1724 +#include "union.h"
1725 +
1726 +/*
1727 + * For detailed explanation of copyup see:
1728 + * Documentation/filesystems/unionfs/concepts.txt
1729 + */
1730 +
1731 +#ifdef CONFIG_UNION_FS_XATTR
1732 +/* copyup all extended attrs for a given dentry */
1733 +static int copyup_xattrs(struct dentry *old_lower_dentry,
1734 + struct dentry *new_lower_dentry)
1735 +{
1736 + int err = 0;
1737 + ssize_t list_size = -1;
1738 + char *name_list = NULL;
1739 + char *attr_value = NULL;
1740 + char *name_list_buf = NULL;
1741 +
1742 + /* query the actual size of the xattr list */
1743 + list_size = vfs_listxattr(old_lower_dentry, NULL, 0);
1744 + if (list_size <= 0) {
1745 + err = list_size;
1746 + goto out;
1747 + }
1748 +
1749 + /* allocate space for the actual list */
1750 + name_list = unionfs_xattr_alloc(list_size + 1, XATTR_LIST_MAX);
1751 + if (unlikely(!name_list || IS_ERR(name_list))) {
1752 + err = PTR_ERR(name_list);
1753 + goto out;
1754 + }
1755 +
1756 + name_list_buf = name_list; /* save for kfree at end */
1757 +
1758 + /* now get the actual xattr list of the source file */
1759 + list_size = vfs_listxattr(old_lower_dentry, name_list, list_size);
1760 + if (list_size <= 0) {
1761 + err = list_size;
1762 + goto out;
1763 + }
1764 +
1765 + /* allocate space to hold each xattr's value */
1766 + attr_value = unionfs_xattr_alloc(XATTR_SIZE_MAX, XATTR_SIZE_MAX);
1767 + if (unlikely(!attr_value || IS_ERR(attr_value))) {
1768 + err = PTR_ERR(name_list);
1769 + goto out;
1770 + }
1771 +
1772 + /* in a loop, get and set each xattr from src to dst file */
1773 + while (*name_list) {
1774 + ssize_t size;
1775 +
1776 + /* Lock here since vfs_getxattr doesn't lock for us */
1777 + mutex_lock(&old_lower_dentry->d_inode->i_mutex);
1778 + size = vfs_getxattr(old_lower_dentry, name_list,
1779 + attr_value, XATTR_SIZE_MAX);
1780 + mutex_unlock(&old_lower_dentry->d_inode->i_mutex);
1781 + if (size < 0) {
1782 + err = size;
1783 + goto out;
1784 + }
1785 + if (size > XATTR_SIZE_MAX) {
1786 + err = -E2BIG;
1787 + goto out;
1788 + }
1789 + /* Don't lock here since vfs_setxattr does it for us. */
1790 + err = vfs_setxattr(new_lower_dentry, name_list, attr_value,
1791 + size, 0);
1792 + /*
1793 + * Selinux depends on "security.*" xattrs, so to maintain
1794 + * the security of copied-up files, if Selinux is active,
1795 + * then we must copy these xattrs as well. So we need to
1796 + * temporarily get FOWNER privileges.
1797 + * XXX: move entire copyup code to SIOQ.
1798 + */
1799 + if (err == -EPERM && !capable(CAP_FOWNER)) {
1800 + const struct cred *old_creds;
1801 + struct cred *new_creds;
1802 +
1803 + new_creds = prepare_creds();
1804 + if (unlikely(!new_creds)) {
1805 + err = -ENOMEM;
1806 + goto out;
1807 + }
1808 + cap_raise(new_creds->cap_effective, CAP_FOWNER);
1809 + old_creds = override_creds(new_creds);
1810 + err = vfs_setxattr(new_lower_dentry, name_list,
1811 + attr_value, size, 0);
1812 + revert_creds(old_creds);
1813 + }
1814 + if (err < 0)
1815 + goto out;
1816 + name_list += strlen(name_list) + 1;
1817 + }
1818 +out:
1819 + unionfs_xattr_kfree(name_list_buf);
1820 + unionfs_xattr_kfree(attr_value);
1821 + /* Ignore if xattr isn't supported */
1822 + if (err == -ENOTSUPP || err == -EOPNOTSUPP)
1823 + err = 0;
1824 + return err;
1825 +}
1826 +#endif /* CONFIG_UNION_FS_XATTR */
1827 +
1828 +/*
1829 + * Determine the mode based on the copyup flags, and the existing dentry.
1830 + *
1831 + * Handle file systems which may not support certain options. For example
1832 + * jffs2 doesn't allow one to chmod a symlink. So we ignore such harmless
1833 + * errors, rather than propagating them up, which results in copyup errors
1834 + * and errors returned back to users.
1835 + */
1836 +static int copyup_permissions(struct super_block *sb,
1837 + struct dentry *old_lower_dentry,
1838 + struct dentry *new_lower_dentry)
1839 +{
1840 + struct inode *i = old_lower_dentry->d_inode;
1841 + struct iattr newattrs;
1842 + int err;
1843 +
1844 + newattrs.ia_atime = i->i_atime;
1845 + newattrs.ia_mtime = i->i_mtime;
1846 + newattrs.ia_ctime = i->i_ctime;
1847 + newattrs.ia_gid = i->i_gid;
1848 + newattrs.ia_uid = i->i_uid;
1849 + newattrs.ia_valid = ATTR_CTIME | ATTR_ATIME | ATTR_MTIME |
1850 + ATTR_ATIME_SET | ATTR_MTIME_SET | ATTR_FORCE |
1851 + ATTR_GID | ATTR_UID;
1852 + mutex_lock(&new_lower_dentry->d_inode->i_mutex);
1853 + err = notify_change(new_lower_dentry, &newattrs);
1854 + if (err)
1855 + goto out;
1856 +
1857 + /* now try to change the mode and ignore EOPNOTSUPP on symlinks */
1858 + newattrs.ia_mode = i->i_mode;
1859 + newattrs.ia_valid = ATTR_MODE | ATTR_FORCE;
1860 + err = notify_change(new_lower_dentry, &newattrs);
1861 + if (err == -EOPNOTSUPP &&
1862 + S_ISLNK(new_lower_dentry->d_inode->i_mode)) {
1863 + printk(KERN_WARNING
1864 + "unionfs: changing \"%s\" symlink mode unsupported\n",
1865 + new_lower_dentry->d_name.name);
1866 + err = 0;
1867 + }
1868 +
1869 +out:
1870 + mutex_unlock(&new_lower_dentry->d_inode->i_mutex);
1871 + return err;
1872 +}
1873 +
1874 +/*
1875 + * create the new device/file/directory - use copyup_permission to copyup
1876 + * times, and mode
1877 + *
1878 + * if the object being copied up is a regular file, the file is only created,
1879 + * the contents have to be copied up separately
1880 + */
1881 +static int __copyup_ndentry(struct dentry *old_lower_dentry,
1882 + struct dentry *new_lower_dentry,
1883 + struct dentry *new_lower_parent_dentry,
1884 + char *symbuf)
1885 +{
1886 + int err = 0;
1887 + umode_t old_mode = old_lower_dentry->d_inode->i_mode;
1888 + struct sioq_args args;
1889 +
1890 + if (S_ISDIR(old_mode)) {
1891 + args.mkdir.parent = new_lower_parent_dentry->d_inode;
1892 + args.mkdir.dentry = new_lower_dentry;
1893 + args.mkdir.mode = old_mode;
1894 +
1895 + run_sioq(__unionfs_mkdir, &args);
1896 + err = args.err;
1897 + } else if (S_ISLNK(old_mode)) {
1898 + args.symlink.parent = new_lower_parent_dentry->d_inode;
1899 + args.symlink.dentry = new_lower_dentry;
1900 + args.symlink.symbuf = symbuf;
1901 +
1902 + run_sioq(__unionfs_symlink, &args);
1903 + err = args.err;
1904 + } else if (S_ISBLK(old_mode) || S_ISCHR(old_mode) ||
1905 + S_ISFIFO(old_mode) || S_ISSOCK(old_mode)) {
1906 + args.mknod.parent = new_lower_parent_dentry->d_inode;
1907 + args.mknod.dentry = new_lower_dentry;
1908 + args.mknod.mode = old_mode;
1909 + args.mknod.dev = old_lower_dentry->d_inode->i_rdev;
1910 +
1911 + run_sioq(__unionfs_mknod, &args);
1912 + err = args.err;
1913 + } else if (S_ISREG(old_mode)) {
1914 + struct nameidata nd;
1915 + err = init_lower_nd(&nd, LOOKUP_CREATE);
1916 + if (unlikely(err < 0))
1917 + goto out;
1918 + args.create.nd = &nd;
1919 + args.create.parent = new_lower_parent_dentry->d_inode;
1920 + args.create.dentry = new_lower_dentry;
1921 + args.create.mode = old_mode;
1922 +
1923 + run_sioq(__unionfs_create, &args);
1924 + err = args.err;
1925 + release_lower_nd(&nd, err);
1926 + } else {
1927 + printk(KERN_CRIT "unionfs: unknown inode type %d\n",
1928 + old_mode);
1929 + BUG();
1930 + }
1931 +
1932 +out:
1933 + return err;
1934 +}
1935 +
1936 +static int __copyup_reg_data(struct dentry *dentry,
1937 + struct dentry *new_lower_dentry, int new_bindex,
1938 + struct dentry *old_lower_dentry, int old_bindex,
1939 + struct file **copyup_file, loff_t len)
1940 +{
1941 + struct super_block *sb = dentry->d_sb;
1942 + struct file *input_file;
1943 + struct file *output_file;
1944 + struct vfsmount *output_mnt;
1945 + mm_segment_t old_fs;
1946 + char *buf = NULL;
1947 + ssize_t read_bytes, write_bytes;
1948 + loff_t size;
1949 + int err = 0;
1950 +
1951 + /* open old file */
1952 + unionfs_mntget(dentry, old_bindex);
1953 + branchget(sb, old_bindex);
1954 + /* dentry_open calls dput and mntput if it returns an error */
1955 + input_file = dentry_open(old_lower_dentry,
1956 + unionfs_lower_mnt_idx(dentry, old_bindex),
1957 + O_RDONLY | O_LARGEFILE, current_cred());
1958 + if (IS_ERR(input_file)) {
1959 + dput(old_lower_dentry);
1960 + err = PTR_ERR(input_file);
1961 + goto out;
1962 + }
1963 + if (unlikely(!input_file->f_op || !input_file->f_op->read)) {
1964 + err = -EINVAL;
1965 + goto out_close_in;
1966 + }
1967 +
1968 + /* open new file */
1969 + dget(new_lower_dentry);
1970 + output_mnt = unionfs_mntget(sb->s_root, new_bindex);
1971 + branchget(sb, new_bindex);
1972 + output_file = dentry_open(new_lower_dentry, output_mnt,
1973 + O_RDWR | O_LARGEFILE, current_cred());
1974 + if (IS_ERR(output_file)) {
1975 + err = PTR_ERR(output_file);
1976 + goto out_close_in2;
1977 + }
1978 + if (unlikely(!output_file->f_op || !output_file->f_op->write)) {
1979 + err = -EINVAL;
1980 + goto out_close_out;
1981 + }
1982 +
1983 + /* allocating a buffer */
1984 + buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
1985 + if (unlikely(!buf)) {
1986 + err = -ENOMEM;
1987 + goto out_close_out;
1988 + }
1989 +
1990 + input_file->f_pos = 0;
1991 + output_file->f_pos = 0;
1992 +
1993 + old_fs = get_fs();
1994 + set_fs(KERNEL_DS);
1995 +
1996 + size = len;
1997 + err = 0;
1998 + do {
1999 + if (len >= PAGE_SIZE)
2000 + size = PAGE_SIZE;
2001 + else if ((len < PAGE_SIZE) && (len > 0))
2002 + size = len;
2003 +
2004 + len -= PAGE_SIZE;
2005 +
2006 + read_bytes =
2007 + input_file->f_op->read(input_file,
2008 + (char __user *)buf, size,
2009 + &input_file->f_pos);
2010 + if (read_bytes <= 0) {
2011 + err = read_bytes;
2012 + break;
2013 + }
2014 +
2015 + /* see Documentation/filesystems/unionfs/issues.txt */
2016 + lockdep_off();
2017 + write_bytes =
2018 + output_file->f_op->write(output_file,
2019 + (char __user *)buf,
2020 + read_bytes,
2021 + &output_file->f_pos);
2022 + lockdep_on();
2023 + if ((write_bytes < 0) || (write_bytes < read_bytes)) {
2024 + err = write_bytes;
2025 + break;
2026 + }
2027 + } while ((read_bytes > 0) && (len > 0));
2028 +
2029 + set_fs(old_fs);
2030 +
2031 + kfree(buf);
2032 +
2033 + if (!err)
2034 + err = output_file->f_op->fsync(output_file,
2035 + new_lower_dentry, 0);
2036 +
2037 + if (err)
2038 + goto out_close_out;
2039 +
2040 + if (copyup_file) {
2041 + *copyup_file = output_file;
2042 + goto out_close_in;
2043 + }
2044 +
2045 +out_close_out:
2046 + fput(output_file);
2047 +
2048 +out_close_in2:
2049 + branchput(sb, new_bindex);
2050 +
2051 +out_close_in:
2052 + fput(input_file);
2053 +
2054 +out:
2055 + branchput(sb, old_bindex);
2056 +
2057 + return err;
2058 +}
2059 +
2060 +/*
2061 + * dput the lower references for old and new dentry & clear a lower dentry
2062 + * pointer
2063 + */
2064 +static void __clear(struct dentry *dentry, struct dentry *old_lower_dentry,
2065 + int old_bstart, int old_bend,
2066 + struct dentry *new_lower_dentry, int new_bindex)
2067 +{
2068 + /* get rid of the lower dentry and all its traces */
2069 + unionfs_set_lower_dentry_idx(dentry, new_bindex, NULL);
2070 + dbstart(dentry) = old_bstart;
2071 + dbend(dentry) = old_bend;
2072 +
2073 + dput(new_lower_dentry);
2074 + dput(old_lower_dentry);
2075 +}
2076 +
2077 +/*
2078 + * Copy up a dentry to a file of specified name.
2079 + *
2080 + * @dir: used to pull the ->i_sb to access other branches
2081 + * @dentry: the non-negative dentry whose lower_inode we should copy
2082 + * @bstart: the branch of the lower_inode to copy from
2083 + * @new_bindex: the branch to create the new file in
2084 + * @name: the name of the file to create
2085 + * @namelen: length of @name
2086 + * @copyup_file: the "struct file" to return (optional)
2087 + * @len: how many bytes to copy-up?
2088 + */
2089 +int copyup_dentry(struct inode *dir, struct dentry *dentry, int bstart,
2090 + int new_bindex, const char *name, int namelen,
2091 + struct file **copyup_file, loff_t len)
2092 +{
2093 + struct dentry *new_lower_dentry;
2094 + struct dentry *old_lower_dentry = NULL;
2095 + struct super_block *sb;
2096 + int err = 0;
2097 + int old_bindex;
2098 + int old_bstart;
2099 + int old_bend;
2100 + struct dentry *new_lower_parent_dentry = NULL;
2101 + mm_segment_t oldfs;
2102 + char *symbuf = NULL;
2103 +
2104 + verify_locked(dentry);
2105 +
2106 + old_bindex = bstart;
2107 + old_bstart = dbstart(dentry);
2108 + old_bend = dbend(dentry);
2109 +
2110 + BUG_ON(new_bindex < 0);
2111 + BUG_ON(new_bindex >= old_bindex);
2112 +
2113 + sb = dir->i_sb;
2114 +
2115 + err = is_robranch_super(sb, new_bindex);
2116 + if (err)
2117 + goto out;
2118 +
2119 + /* Create the directory structure above this dentry. */
2120 + new_lower_dentry = create_parents(dir, dentry, name, new_bindex);
2121 + if (IS_ERR(new_lower_dentry)) {
2122 + err = PTR_ERR(new_lower_dentry);
2123 + goto out;
2124 + }
2125 +
2126 + old_lower_dentry = unionfs_lower_dentry_idx(dentry, old_bindex);
2127 + /* we conditionally dput this old_lower_dentry at end of function */
2128 + dget(old_lower_dentry);
2129 +
2130 + /* For symlinks, we must read the link before we lock the directory. */
2131 + if (S_ISLNK(old_lower_dentry->d_inode->i_mode)) {
2132 +
2133 + symbuf = kmalloc(PATH_MAX, GFP_KERNEL);
2134 + if (unlikely(!symbuf)) {
2135 + __clear(dentry, old_lower_dentry,
2136 + old_bstart, old_bend,
2137 + new_lower_dentry, new_bindex);
2138 + err = -ENOMEM;
2139 + goto out_free;
2140 + }
2141 +
2142 + oldfs = get_fs();
2143 + set_fs(KERNEL_DS);
2144 + err = old_lower_dentry->d_inode->i_op->readlink(
2145 + old_lower_dentry,
2146 + (char __user *)symbuf,
2147 + PATH_MAX);
2148 + set_fs(oldfs);
2149 + if (err < 0) {
2150 + __clear(dentry, old_lower_dentry,
2151 + old_bstart, old_bend,
2152 + new_lower_dentry, new_bindex);
2153 + goto out_free;
2154 + }
2155 + symbuf[err] = '\0';
2156 + }
2157 +
2158 + /* Now we lock the parent, and create the object in the new branch. */
2159 + new_lower_parent_dentry = lock_parent(new_lower_dentry);
2160 +
2161 + /* create the new inode */
2162 + err = __copyup_ndentry(old_lower_dentry, new_lower_dentry,
2163 + new_lower_parent_dentry, symbuf);
2164 +
2165 + if (err) {
2166 + __clear(dentry, old_lower_dentry,
2167 + old_bstart, old_bend,
2168 + new_lower_dentry, new_bindex);
2169 + goto out_unlock;
2170 + }
2171 +
2172 + /* We actually copyup the file here. */
2173 + if (S_ISREG(old_lower_dentry->d_inode->i_mode))
2174 + err = __copyup_reg_data(dentry, new_lower_dentry, new_bindex,
2175 + old_lower_dentry, old_bindex,
2176 + copyup_file, len);
2177 + if (err)
2178 + goto out_unlink;
2179 +
2180 + /* Set permissions. */
2181 + err = copyup_permissions(sb, old_lower_dentry, new_lower_dentry);
2182 + if (err)
2183 + goto out_unlink;
2184 +
2185 +#ifdef CONFIG_UNION_FS_XATTR
2186 + /* Selinux uses extended attributes for permissions. */
2187 + err = copyup_xattrs(old_lower_dentry, new_lower_dentry);
2188 + if (err)
2189 + goto out_unlink;
2190 +#endif /* CONFIG_UNION_FS_XATTR */
2191 +
2192 + /* do not allow files getting deleted to be re-interposed */
2193 + if (!d_deleted(dentry))
2194 + unionfs_reinterpose(dentry);
2195 +
2196 + goto out_unlock;
2197 +
2198 +out_unlink:
2199 + /*
2200 + * copyup failed, because we possibly ran out of space or
2201 + * quota, or something else happened so let's unlink; we don't
2202 + * really care about the return value of vfs_unlink
2203 + */
2204 + vfs_unlink(new_lower_parent_dentry->d_inode, new_lower_dentry);
2205 +
2206 + if (copyup_file) {
2207 + /* need to close the file */
2208 +
2209 + fput(*copyup_file);
2210 + branchput(sb, new_bindex);
2211 + }
2212 +
2213 + /*
2214 + * TODO: should we reset the error to something like -EIO?
2215 + *
2216 + * If we don't reset, the user may get some nonsensical errors, but
2217 + * on the other hand, if we reset to EIO, we guarantee that the user
2218 + * will get a "confusing" error message.
2219 + */
2220 +
2221 +out_unlock:
2222 + unlock_dir(new_lower_parent_dentry);
2223 +
2224 +out_free:
2225 + /*
2226 + * If old_lower_dentry was not a file, then we need to dput it. If
2227 + * it was a file, then it was already dput indirectly by other
2228 + * functions we call above which operate on regular files.
2229 + */
2230 + if (old_lower_dentry && old_lower_dentry->d_inode &&
2231 + !S_ISREG(old_lower_dentry->d_inode->i_mode))
2232 + dput(old_lower_dentry);
2233 + kfree(symbuf);
2234 +
2235 + if (err) {
2236 + /*
2237 + * if directory creation succeeded, but inode copyup failed,
2238 + * then purge new dentries.
2239 + */
2240 + if (dbstart(dentry) < old_bstart &&
2241 + ibstart(dentry->d_inode) > dbstart(dentry))
2242 + __clear(dentry, NULL, old_bstart, old_bend,
2243 + unionfs_lower_dentry(dentry), dbstart(dentry));
2244 + goto out;
2245 + }
2246 + if (!S_ISDIR(dentry->d_inode->i_mode)) {
2247 + unionfs_postcopyup_release(dentry);
2248 + if (!unionfs_lower_inode(dentry->d_inode)) {
2249 + /*
2250 + * If we got here, then we copied up to an
2251 + * unlinked-open file, whose name is .unionfsXXXXX.
2252 + */
2253 + struct inode *inode = new_lower_dentry->d_inode;
2254 + atomic_inc(&inode->i_count);
2255 + unionfs_set_lower_inode_idx(dentry->d_inode,
2256 + ibstart(dentry->d_inode),
2257 + inode);
2258 + }
2259 + }
2260 + unionfs_postcopyup_setmnt(dentry);
2261 + /* sync inode times from copied-up inode to our inode */
2262 + unionfs_copy_attr_times(dentry->d_inode);
2263 + unionfs_check_inode(dir);
2264 + unionfs_check_dentry(dentry);
2265 +out:
2266 + return err;
2267 +}
2268 +
2269 +/*
2270 + * This function creates a copy of a file represented by 'file' which
2271 + * currently resides in branch 'bstart' to branch 'new_bindex.' The copy
2272 + * will be named "name".
2273 + */
2274 +int copyup_named_file(struct inode *dir, struct file *file, char *name,
2275 + int bstart, int new_bindex, loff_t len)
2276 +{
2277 + int err = 0;
2278 + struct file *output_file = NULL;
2279 +
2280 + err = copyup_dentry(dir, file->f_path.dentry, bstart, new_bindex,
2281 + name, strlen(name), &output_file, len);
2282 + if (!err) {
2283 + fbstart(file) = new_bindex;
2284 + unionfs_set_lower_file_idx(file, new_bindex, output_file);
2285 + }
2286 +
2287 + return err;
2288 +}
2289 +
2290 +/*
2291 + * This function creates a copy of a file represented by 'file' which
2292 + * currently resides in branch 'bstart' to branch 'new_bindex'.
2293 + */
2294 +int copyup_file(struct inode *dir, struct file *file, int bstart,
2295 + int new_bindex, loff_t len)
2296 +{
2297 + int err = 0;
2298 + struct file *output_file = NULL;
2299 + struct dentry *dentry = file->f_path.dentry;
2300 +
2301 + err = copyup_dentry(dir, dentry, bstart, new_bindex,
2302 + dentry->d_name.name, dentry->d_name.len,
2303 + &output_file, len);
2304 + if (!err) {
2305 + fbstart(file) = new_bindex;
2306 + unionfs_set_lower_file_idx(file, new_bindex, output_file);
2307 + }
2308 +
2309 + return err;
2310 +}
2311 +
2312 +/* purge a dentry's lower-branch states (dput/mntput, etc.) */
2313 +static void __cleanup_dentry(struct dentry *dentry, int bindex,
2314 + int old_bstart, int old_bend)
2315 +{
2316 + int loop_start;
2317 + int loop_end;
2318 + int new_bstart = -1;
2319 + int new_bend = -1;
2320 + int i;
2321 +
2322 + loop_start = min(old_bstart, bindex);
2323 + loop_end = max(old_bend, bindex);
2324 +
2325 + /*
2326 + * This loop sets the bstart and bend for the new dentry by
2327 + * traversing from left to right. It also dputs all negative
2328 + * dentries except bindex
2329 + */
2330 + for (i = loop_start; i <= loop_end; i++) {
2331 + if (!unionfs_lower_dentry_idx(dentry, i))
2332 + continue;
2333 +
2334 + if (i == bindex) {
2335 + new_bend = i;
2336 + if (new_bstart < 0)
2337 + new_bstart = i;
2338 + continue;
2339 + }
2340 +
2341 + if (!unionfs_lower_dentry_idx(dentry, i)->d_inode) {
2342 + dput(unionfs_lower_dentry_idx(dentry, i));
2343 + unionfs_set_lower_dentry_idx(dentry, i, NULL);
2344 +
2345 + unionfs_mntput(dentry, i);
2346 + unionfs_set_lower_mnt_idx(dentry, i, NULL);
2347 + } else {
2348 + if (new_bstart < 0)
2349 + new_bstart = i;
2350 + new_bend = i;
2351 + }
2352 + }
2353 +
2354 + if (new_bstart < 0)
2355 + new_bstart = bindex;
2356 + if (new_bend < 0)
2357 + new_bend = bindex;
2358 + dbstart(dentry) = new_bstart;
2359 + dbend(dentry) = new_bend;
2360 +
2361 +}
2362 +
2363 +/* set lower inode ptr and update bstart & bend if necessary */
2364 +static void __set_inode(struct dentry *upper, struct dentry *lower,
2365 + int bindex)
2366 +{
2367 + unionfs_set_lower_inode_idx(upper->d_inode, bindex,
2368 + igrab(lower->d_inode));
2369 + if (likely(ibstart(upper->d_inode) > bindex))
2370 + ibstart(upper->d_inode) = bindex;
2371 + if (likely(ibend(upper->d_inode) < bindex))
2372 + ibend(upper->d_inode) = bindex;
2373 +
2374 +}
2375 +
2376 +/* set lower dentry ptr and update bstart & bend if necessary */
2377 +static void __set_dentry(struct dentry *upper, struct dentry *lower,
2378 + int bindex)
2379 +{
2380 + unionfs_set_lower_dentry_idx(upper, bindex, lower);
2381 + if (likely(dbstart(upper) > bindex))
2382 + dbstart(upper) = bindex;
2383 + if (likely(dbend(upper) < bindex))
2384 + dbend(upper) = bindex;
2385 +}
2386 +
2387 +/*
2388 + * This function replicates the directory structure up-to given dentry
2389 + * in the bindex branch.
2390 + */
2391 +struct dentry *create_parents(struct inode *dir, struct dentry *dentry,
2392 + const char *name, int bindex)
2393 +{
2394 + int err;
2395 + struct dentry *child_dentry;
2396 + struct dentry *parent_dentry;
2397 + struct dentry *lower_parent_dentry = NULL;
2398 + struct dentry *lower_dentry = NULL;
2399 + const char *childname;
2400 + unsigned int childnamelen;
2401 + int nr_dentry;
2402 + int count = 0;
2403 + int old_bstart;
2404 + int old_bend;
2405 + struct dentry **path = NULL;
2406 + struct super_block *sb;
2407 +
2408 + verify_locked(dentry);
2409 +
2410 + err = is_robranch_super(dir->i_sb, bindex);
2411 + if (err) {
2412 + lower_dentry = ERR_PTR(err);
2413 + goto out;
2414 + }
2415 +
2416 + old_bstart = dbstart(dentry);
2417 + old_bend = dbend(dentry);
2418 +
2419 + lower_dentry = ERR_PTR(-ENOMEM);
2420 +
2421 + /* There is no sense allocating any less than the minimum. */
2422 + nr_dentry = 1;
2423 + path = kmalloc(nr_dentry * sizeof(struct dentry *), GFP_KERNEL);
2424 + if (unlikely(!path))
2425 + goto out;
2426 +
2427 + /* assume the negative dentry of unionfs as the parent dentry */
2428 + parent_dentry = dentry;
2429 +
2430 + /*
2431 + * This loop finds the first parent that exists in the given branch.
2432 + * We start building the directory structure from there. At the end
2433 + * of the loop, the following should hold:
2434 + * - child_dentry is the first nonexistent child
2435 + * - parent_dentry is the first existent parent
2436 + * - path[0] is the = deepest child
2437 + * - path[count] is the first child to create
2438 + */
2439 + do {
2440 + child_dentry = parent_dentry;
2441 +
2442 + /* find the parent directory dentry in unionfs */
2443 + parent_dentry = dget_parent(child_dentry);
2444 +
2445 + /* find out the lower_parent_dentry in the given branch */
2446 + lower_parent_dentry =
2447 + unionfs_lower_dentry_idx(parent_dentry, bindex);
2448 +
2449 + /* grow path table */
2450 + if (count == nr_dentry) {
2451 + void *p;
2452 +
2453 + nr_dentry *= 2;
2454 + p = krealloc(path, nr_dentry * sizeof(struct dentry *),
2455 + GFP_KERNEL);
2456 + if (unlikely(!p)) {
2457 + lower_dentry = ERR_PTR(-ENOMEM);
2458 + goto out;
2459 + }
2460 + path = p;
2461 + }
2462 +
2463 + /* store the child dentry */
2464 + path[count++] = child_dentry;
2465 + } while (!lower_parent_dentry);
2466 + count--;
2467 +
2468 + sb = dentry->d_sb;
2469 +
2470 + /*
2471 + * This code goes between the begin/end labels and basically
2472 + * emulates a while(child_dentry != dentry), only cleaner and
2473 + * shorter than what would be a much longer while loop.
2474 + */
2475 +begin:
2476 + /* get lower parent dir in the current branch */
2477 + lower_parent_dentry = unionfs_lower_dentry_idx(parent_dentry, bindex);
2478 + dput(parent_dentry);
2479 +
2480 + /* init the values to lookup */
2481 + childname = child_dentry->d_name.name;
2482 + childnamelen = child_dentry->d_name.len;
2483 +
2484 + if (child_dentry != dentry) {
2485 + /* lookup child in the underlying file system */
2486 + lower_dentry = lookup_one_len(childname, lower_parent_dentry,
2487 + childnamelen);
2488 + if (IS_ERR(lower_dentry))
2489 + goto out;
2490 + } else {
2491 + /*
2492 + * Is the name a whiteout of the child name ? lookup the
2493 + * whiteout child in the underlying file system
2494 + */
2495 + lower_dentry = lookup_one_len(name, lower_parent_dentry,
2496 + strlen(name));
2497 + if (IS_ERR(lower_dentry))
2498 + goto out;
2499 +
2500 + /* Replace the current dentry (if any) with the new one */
2501 + dput(unionfs_lower_dentry_idx(dentry, bindex));
2502 + unionfs_set_lower_dentry_idx(dentry, bindex,
2503 + lower_dentry);
2504 +
2505 + __cleanup_dentry(dentry, bindex, old_bstart, old_bend);
2506 + goto out;
2507 + }
2508 +
2509 + if (lower_dentry->d_inode) {
2510 + /*
2511 + * since this already exists we dput to avoid
2512 + * multiple references on the same dentry
2513 + */
2514 + dput(lower_dentry);
2515 + } else {
2516 + struct sioq_args args;
2517 +
2518 + /* it's a negative dentry, create a new dir */
2519 + lower_parent_dentry = lock_parent(lower_dentry);
2520 +
2521 + args.mkdir.parent = lower_parent_dentry->d_inode;
2522 + args.mkdir.dentry = lower_dentry;
2523 + args.mkdir.mode = child_dentry->d_inode->i_mode;
2524 +
2525 + run_sioq(__unionfs_mkdir, &args);
2526 + err = args.err;
2527 +
2528 + if (!err)
2529 + err = copyup_permissions(dir->i_sb, child_dentry,
2530 + lower_dentry);
2531 + unlock_dir(lower_parent_dentry);
2532 + if (err) {
2533 + dput(lower_dentry);
2534 + lower_dentry = ERR_PTR(err);
2535 + goto out;
2536 + }
2537 +
2538 + }
2539 +
2540 + __set_inode(child_dentry, lower_dentry, bindex);
2541 + __set_dentry(child_dentry, lower_dentry, bindex);
2542 + /*
2543 + * update times of this dentry, but also the parent, because if
2544 + * we changed, the parent may have changed too.
2545 + */
2546 + fsstack_copy_attr_times(parent_dentry->d_inode,
2547 + lower_parent_dentry->d_inode);
2548 + unionfs_copy_attr_times(child_dentry->d_inode);
2549 +
2550 + parent_dentry = child_dentry;
2551 + child_dentry = path[--count];
2552 + goto begin;
2553 +out:
2554 + /* cleanup any leftover locks from the do/while loop above */
2555 + if (IS_ERR(lower_dentry))
2556 + while (count)
2557 + dput(path[count--]);
2558 + kfree(path);
2559 + return lower_dentry;
2560 +}
2561 +
2562 +/*
2563 + * Post-copyup helper to ensure we have valid mnts: set lower mnt of
2564 + * dentry+parents to the first parent node that has an mnt.
2565 + */
2566 +void unionfs_postcopyup_setmnt(struct dentry *dentry)
2567 +{
2568 + struct dentry *parent, *hasone;
2569 + int bindex = dbstart(dentry);
2570 +
2571 + if (unionfs_lower_mnt_idx(dentry, bindex))
2572 + return;
2573 + hasone = dentry->d_parent;
2574 + /* this loop should stop at root dentry */
2575 + while (!unionfs_lower_mnt_idx(hasone, bindex))
2576 + hasone = hasone->d_parent;
2577 + parent = dentry;
2578 + while (!unionfs_lower_mnt_idx(parent, bindex)) {
2579 + unionfs_set_lower_mnt_idx(parent, bindex,
2580 + unionfs_mntget(hasone, bindex));
2581 + parent = parent->d_parent;
2582 + }
2583 +}
2584 +
2585 +/*
2586 + * Post-copyup helper to release all non-directory source objects of a
2587 + * copied-up file. Regular files should have only one lower object.
2588 + */
2589 +void unionfs_postcopyup_release(struct dentry *dentry)
2590 +{
2591 + int bstart, bend;
2592 +
2593 + BUG_ON(S_ISDIR(dentry->d_inode->i_mode));
2594 + bstart = dbstart(dentry);
2595 + bend = dbend(dentry);
2596 +
2597 + path_put_lowers(dentry, bstart + 1, bend, false);
2598 + iput_lowers(dentry->d_inode, bstart + 1, bend, false);
2599 +
2600 + dbend(dentry) = bstart;
2601 + ibend(dentry->d_inode) = ibstart(dentry->d_inode) = bstart;
2602 +}
2603 diff -Naur linux-2.6.30/fs/unionfs/debug.c linux-2.6.30-magellan/fs/unionfs/debug.c
2604 --- linux-2.6.30/fs/unionfs/debug.c 1970-01-01 01:00:00.000000000 +0100
2605 +++ linux-2.6.30-magellan/fs/unionfs/debug.c 2009-06-11 18:39:19.000000000 +0200
2606 @@ -0,0 +1,533 @@
2607 +/*
2608 + * Copyright (c) 2003-2009 Erez Zadok
2609 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
2610 + * Copyright (c) 2003-2009 Stony Brook University
2611 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
2612 + *
2613 + * This program is free software; you can redistribute it and/or modify
2614 + * it under the terms of the GNU General Public License version 2 as
2615 + * published by the Free Software Foundation.
2616 + */
2617 +
2618 +#include "union.h"
2619 +
2620 +/*
2621 + * Helper debugging functions for maintainers (and for users to report back
2622 + * useful information back to maintainers)
2623 + */
2624 +
2625 +/* it's always useful to know what part of the code called us */
2626 +#define PRINT_CALLER(fname, fxn, line) \
2627 + do { \
2628 + if (!printed_caller) { \
2629 + pr_debug("PC:%s:%s:%d\n", (fname), (fxn), (line)); \
2630 + printed_caller = 1; \
2631 + } \
2632 + } while (0)
2633 +
2634 +/*
2635 + * __unionfs_check_{inode,dentry,file} perform exhaustive sanity checking on
2636 + * the fan-out of various Unionfs objects. We check that no lower objects
2637 + * exist outside the start/end branch range; that all objects within are
2638 + * non-NULL (with some allowed exceptions); that for every lower file
2639 + * there's a lower dentry+inode; that the start/end ranges match for all
2640 + * corresponding lower objects; that open files/symlinks have only one lower
2641 + * objects, but directories can have several; and more.
2642 + */
2643 +void __unionfs_check_inode(const struct inode *inode,
2644 + const char *fname, const char *fxn, int line)
2645 +{
2646 + int bindex;
2647 + int istart, iend;
2648 + struct inode *lower_inode;
2649 + struct super_block *sb;
2650 + int printed_caller = 0;
2651 + void *poison_ptr;
2652 +
2653 + /* for inodes now */
2654 + BUG_ON(!inode);
2655 + sb = inode->i_sb;
2656 + istart = ibstart(inode);
2657 + iend = ibend(inode);
2658 + /* don't check inode if no lower branches */
2659 + if (istart < 0 && iend < 0)
2660 + return;
2661 + if (unlikely(istart > iend)) {
2662 + PRINT_CALLER(fname, fxn, line);
2663 + pr_debug(" Ci0: inode=%p istart/end=%d:%d\n",
2664 + inode, istart, iend);
2665 + }
2666 + if (unlikely((istart == -1 && iend != -1) ||
2667 + (istart != -1 && iend == -1))) {
2668 + PRINT_CALLER(fname, fxn, line);
2669 + pr_debug(" Ci1: inode=%p istart/end=%d:%d\n",
2670 + inode, istart, iend);
2671 + }
2672 + if (!S_ISDIR(inode->i_mode)) {
2673 + if (unlikely(iend != istart)) {
2674 + PRINT_CALLER(fname, fxn, line);
2675 + pr_debug(" Ci2: inode=%p istart=%d iend=%d\n",
2676 + inode, istart, iend);
2677 + }
2678 + }
2679 +
2680 + for (bindex = sbstart(sb); bindex < sbmax(sb); bindex++) {
2681 + if (unlikely(!UNIONFS_I(inode))) {
2682 + PRINT_CALLER(fname, fxn, line);
2683 + pr_debug(" Ci3: no inode_info %p\n", inode);
2684 + return;
2685 + }
2686 + if (unlikely(!UNIONFS_I(inode)->lower_inodes)) {
2687 + PRINT_CALLER(fname, fxn, line);
2688 + pr_debug(" Ci4: no lower_inodes %p\n", inode);
2689 + return;
2690 + }
2691 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
2692 + if (lower_inode) {
2693 + memset(&poison_ptr, POISON_INUSE, sizeof(void *));
2694 + if (unlikely(bindex < istart || bindex > iend)) {
2695 + PRINT_CALLER(fname, fxn, line);
2696 + pr_debug(" Ci5: inode/linode=%p:%p bindex=%d "
2697 + "istart/end=%d:%d\n", inode,
2698 + lower_inode, bindex, istart, iend);
2699 + } else if (unlikely(lower_inode == poison_ptr)) {
2700 + /* freed inode! */
2701 + PRINT_CALLER(fname, fxn, line);
2702 + pr_debug(" Ci6: inode/linode=%p:%p bindex=%d "
2703 + "istart/end=%d:%d\n", inode,
2704 + lower_inode, bindex, istart, iend);
2705 + }
2706 + continue;
2707 + }
2708 + /* if we get here, then lower_inode == NULL */
2709 + if (bindex < istart || bindex > iend)
2710 + continue;
2711 + /*
2712 + * directories can have NULL lower inodes in b/t start/end,
2713 + * but NOT if at the start/end range.
2714 + */
2715 + if (unlikely(S_ISDIR(inode->i_mode) &&
2716 + bindex > istart && bindex < iend))
2717 + continue;
2718 + PRINT_CALLER(fname, fxn, line);
2719 + pr_debug(" Ci7: inode/linode=%p:%p "
2720 + "bindex=%d istart/end=%d:%d\n",
2721 + inode, lower_inode, bindex, istart, iend);
2722 + }
2723 +}
2724 +
2725 +void __unionfs_check_dentry(const struct dentry *dentry,
2726 + const char *fname, const char *fxn, int line)
2727 +{
2728 + int bindex;
2729 + int dstart, dend, istart, iend;
2730 + struct dentry *lower_dentry;
2731 + struct inode *inode, *lower_inode;
2732 + struct super_block *sb;
2733 + struct vfsmount *lower_mnt;
2734 + int printed_caller = 0;
2735 + void *poison_ptr;
2736 +
2737 + BUG_ON(!dentry);
2738 + sb = dentry->d_sb;
2739 + inode = dentry->d_inode;
2740 + dstart = dbstart(dentry);
2741 + dend = dbend(dentry);
2742 + /* don't check dentry/mnt if no lower branches */
2743 + if (dstart < 0 && dend < 0)
2744 + goto check_inode;
2745 + BUG_ON(dstart > dend);
2746 +
2747 + if (unlikely((dstart == -1 && dend != -1) ||
2748 + (dstart != -1 && dend == -1))) {
2749 + PRINT_CALLER(fname, fxn, line);
2750 + pr_debug(" CD0: dentry=%p dstart/end=%d:%d\n",
2751 + dentry, dstart, dend);
2752 + }
2753 + /*
2754 + * check for NULL dentries inside the start/end range, or
2755 + * non-NULL dentries outside the start/end range.
2756 + */
2757 + for (bindex = sbstart(sb); bindex < sbmax(sb); bindex++) {
2758 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
2759 + if (lower_dentry) {
2760 + if (unlikely(bindex < dstart || bindex > dend)) {
2761 + PRINT_CALLER(fname, fxn, line);
2762 + pr_debug(" CD1: dentry/lower=%p:%p(%p) "
2763 + "bindex=%d dstart/end=%d:%d\n",
2764 + dentry, lower_dentry,
2765 + (lower_dentry ? lower_dentry->d_inode :
2766 + (void *) -1L),
2767 + bindex, dstart, dend);
2768 + }
2769 + } else { /* lower_dentry == NULL */
2770 + if (bindex < dstart || bindex > dend)
2771 + continue;
2772 + /*
2773 + * Directories can have NULL lower inodes in b/t
2774 + * start/end, but NOT if at the start/end range.
2775 + * Ignore this rule, however, if this is a NULL
2776 + * dentry or a deleted dentry.
2777 + */
2778 + if (unlikely(!d_deleted((struct dentry *) dentry) &&
2779 + inode &&
2780 + !(inode && S_ISDIR(inode->i_mode) &&
2781 + bindex > dstart && bindex < dend))) {
2782 + PRINT_CALLER(fname, fxn, line);
2783 + pr_debug(" CD2: dentry/lower=%p:%p(%p) "
2784 + "bindex=%d dstart/end=%d:%d\n",
2785 + dentry, lower_dentry,
2786 + (lower_dentry ?
2787 + lower_dentry->d_inode :
2788 + (void *) -1L),
2789 + bindex, dstart, dend);
2790 + }
2791 + }
2792 + }
2793 +
2794 + /* check for vfsmounts same as for dentries */
2795 + for (bindex = sbstart(sb); bindex < sbmax(sb); bindex++) {
2796 + lower_mnt = unionfs_lower_mnt_idx(dentry, bindex);
2797 + if (lower_mnt) {
2798 + if (unlikely(bindex < dstart || bindex > dend)) {
2799 + PRINT_CALLER(fname, fxn, line);
2800 + pr_debug(" CM0: dentry/lmnt=%p:%p bindex=%d "
2801 + "dstart/end=%d:%d\n", dentry,
2802 + lower_mnt, bindex, dstart, dend);
2803 + }
2804 + } else { /* lower_mnt == NULL */
2805 + if (bindex < dstart || bindex > dend)
2806 + continue;
2807 + /*
2808 + * Directories can have NULL lower inodes in b/t
2809 + * start/end, but NOT if at the start/end range.
2810 + * Ignore this rule, however, if this is a NULL
2811 + * dentry.
2812 + */
2813 + if (unlikely(inode &&
2814 + !(inode && S_ISDIR(inode->i_mode) &&
2815 + bindex > dstart && bindex < dend))) {
2816 + PRINT_CALLER(fname, fxn, line);
2817 + pr_debug(" CM1: dentry/lmnt=%p:%p "
2818 + "bindex=%d dstart/end=%d:%d\n",
2819 + dentry, lower_mnt, bindex,
2820 + dstart, dend);
2821 + }
2822 + }
2823 + }
2824 +
2825 +check_inode:
2826 + /* for inodes now */
2827 + if (!inode)
2828 + return;
2829 + istart = ibstart(inode);
2830 + iend = ibend(inode);
2831 + /* don't check inode if no lower branches */
2832 + if (istart < 0 && iend < 0)
2833 + return;
2834 + BUG_ON(istart > iend);
2835 + if (unlikely((istart == -1 && iend != -1) ||
2836 + (istart != -1 && iend == -1))) {
2837 + PRINT_CALLER(fname, fxn, line);
2838 + pr_debug(" CI0: dentry/inode=%p:%p istart/end=%d:%d\n",
2839 + dentry, inode, istart, iend);
2840 + }
2841 + if (unlikely(istart != dstart)) {
2842 + PRINT_CALLER(fname, fxn, line);
2843 + pr_debug(" CI1: dentry/inode=%p:%p istart=%d dstart=%d\n",
2844 + dentry, inode, istart, dstart);
2845 + }
2846 + if (unlikely(iend != dend)) {
2847 + PRINT_CALLER(fname, fxn, line);
2848 + pr_debug(" CI2: dentry/inode=%p:%p iend=%d dend=%d\n",
2849 + dentry, inode, iend, dend);
2850 + }
2851 +
2852 + if (!S_ISDIR(inode->i_mode)) {
2853 + if (unlikely(dend != dstart)) {
2854 + PRINT_CALLER(fname, fxn, line);
2855 + pr_debug(" CI3: dentry/inode=%p:%p dstart=%d dend=%d\n",
2856 + dentry, inode, dstart, dend);
2857 + }
2858 + if (unlikely(iend != istart)) {
2859 + PRINT_CALLER(fname, fxn, line);
2860 + pr_debug(" CI4: dentry/inode=%p:%p istart=%d iend=%d\n",
2861 + dentry, inode, istart, iend);
2862 + }
2863 + }
2864 +
2865 + for (bindex = sbstart(sb); bindex < sbmax(sb); bindex++) {
2866 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
2867 + if (lower_inode) {
2868 + memset(&poison_ptr, POISON_INUSE, sizeof(void *));
2869 + if (unlikely(bindex < istart || bindex > iend)) {
2870 + PRINT_CALLER(fname, fxn, line);
2871 + pr_debug(" CI5: dentry/linode=%p:%p bindex=%d "
2872 + "istart/end=%d:%d\n", dentry,
2873 + lower_inode, bindex, istart, iend);
2874 + } else if (unlikely(lower_inode == poison_ptr)) {
2875 + /* freed inode! */
2876 + PRINT_CALLER(fname, fxn, line);
2877 + pr_debug(" CI6: dentry/linode=%p:%p bindex=%d "
2878 + "istart/end=%d:%d\n", dentry,
2879 + lower_inode, bindex, istart, iend);
2880 + }
2881 + continue;
2882 + }
2883 + /* if we get here, then lower_inode == NULL */
2884 + if (bindex < istart || bindex > iend)
2885 + continue;
2886 + /*
2887 + * directories can have NULL lower inodes in b/t start/end,
2888 + * but NOT if at the start/end range.
2889 + */
2890 + if (unlikely(S_ISDIR(inode->i_mode) &&
2891 + bindex > istart && bindex < iend))
2892 + continue;
2893 + PRINT_CALLER(fname, fxn, line);
2894 + pr_debug(" CI7: dentry/linode=%p:%p "
2895 + "bindex=%d istart/end=%d:%d\n",
2896 + dentry, lower_inode, bindex, istart, iend);
2897 + }
2898 +
2899 + /*
2900 + * If it's a directory, then intermediate objects b/t start/end can
2901 + * be NULL. But, check that all three are NULL: lower dentry, mnt,
2902 + * and inode.
2903 + */
2904 + if (dstart >= 0 && dend >= 0 && S_ISDIR(inode->i_mode))
2905 + for (bindex = dstart+1; bindex < dend; bindex++) {
2906 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
2907 + lower_dentry = unionfs_lower_dentry_idx(dentry,
2908 + bindex);
2909 + lower_mnt = unionfs_lower_mnt_idx(dentry, bindex);
2910 + if (unlikely(!((lower_inode && lower_dentry &&
2911 + lower_mnt) ||
2912 + (!lower_inode &&
2913 + !lower_dentry && !lower_mnt)))) {
2914 + PRINT_CALLER(fname, fxn, line);
2915 + pr_debug(" Cx: lmnt/ldentry/linode=%p:%p:%p "
2916 + "bindex=%d dstart/end=%d:%d\n",
2917 + lower_mnt, lower_dentry, lower_inode,
2918 + bindex, dstart, dend);
2919 + }
2920 + }
2921 + /* check if lower inode is newer than upper one (it shouldn't) */
2922 + if (unlikely(is_newer_lower(dentry) && !is_negative_lower(dentry))) {
2923 + PRINT_CALLER(fname, fxn, line);
2924 + for (bindex = ibstart(inode); bindex <= ibend(inode);
2925 + bindex++) {
2926 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
2927 + if (unlikely(!lower_inode))
2928 + continue;
2929 + pr_debug(" CI8: bindex=%d mtime/lmtime=%lu.%lu/%lu.%lu "
2930 + "ctime/lctime=%lu.%lu/%lu.%lu\n",
2931 + bindex,
2932 + inode->i_mtime.tv_sec,
2933 + inode->i_mtime.tv_nsec,
2934 + lower_inode->i_mtime.tv_sec,
2935 + lower_inode->i_mtime.tv_nsec,
2936 + inode->i_ctime.tv_sec,
2937 + inode->i_ctime.tv_nsec,
2938 + lower_inode->i_ctime.tv_sec,
2939 + lower_inode->i_ctime.tv_nsec);
2940 + }
2941 + }
2942 +}
2943 +
2944 +void __unionfs_check_file(const struct file *file,
2945 + const char *fname, const char *fxn, int line)
2946 +{
2947 + int bindex;
2948 + int dstart, dend, fstart, fend;
2949 + struct dentry *dentry;
2950 + struct file *lower_file;
2951 + struct inode *inode;
2952 + struct super_block *sb;
2953 + int printed_caller = 0;
2954 +
2955 + BUG_ON(!file);
2956 + dentry = file->f_path.dentry;
2957 + sb = dentry->d_sb;
2958 + dstart = dbstart(dentry);
2959 + dend = dbend(dentry);
2960 + BUG_ON(dstart > dend);
2961 + fstart = fbstart(file);
2962 + fend = fbend(file);
2963 + BUG_ON(fstart > fend);
2964 +
2965 + if (unlikely((fstart == -1 && fend != -1) ||
2966 + (fstart != -1 && fend == -1))) {
2967 + PRINT_CALLER(fname, fxn, line);
2968 + pr_debug(" CF0: file/dentry=%p:%p fstart/end=%d:%d\n",
2969 + file, dentry, fstart, fend);
2970 + }
2971 + if (unlikely(fstart != dstart)) {
2972 + PRINT_CALLER(fname, fxn, line);
2973 + pr_debug(" CF1: file/dentry=%p:%p fstart=%d dstart=%d\n",
2974 + file, dentry, fstart, dstart);
2975 + }
2976 + if (unlikely(fend != dend)) {
2977 + PRINT_CALLER(fname, fxn, line);
2978 + pr_debug(" CF2: file/dentry=%p:%p fend=%d dend=%d\n",
2979 + file, dentry, fend, dend);
2980 + }
2981 + inode = dentry->d_inode;
2982 + if (!S_ISDIR(inode->i_mode)) {
2983 + if (unlikely(fend != fstart)) {
2984 + PRINT_CALLER(fname, fxn, line);
2985 + pr_debug(" CF3: file/inode=%p:%p fstart=%d fend=%d\n",
2986 + file, inode, fstart, fend);
2987 + }
2988 + if (unlikely(dend != dstart)) {
2989 + PRINT_CALLER(fname, fxn, line);
2990 + pr_debug(" CF4: file/dentry=%p:%p dstart=%d dend=%d\n",
2991 + file, dentry, dstart, dend);
2992 + }
2993 + }
2994 +
2995 + /*
2996 + * check for NULL dentries inside the start/end range, or
2997 + * non-NULL dentries outside the start/end range.
2998 + */
2999 + for (bindex = sbstart(sb); bindex < sbmax(sb); bindex++) {
3000 + lower_file = unionfs_lower_file_idx(file, bindex);
3001 + if (lower_file) {
3002 + if (unlikely(bindex < fstart || bindex > fend)) {
3003 + PRINT_CALLER(fname, fxn, line);
3004 + pr_debug(" CF5: file/lower=%p:%p bindex=%d "
3005 + "fstart/end=%d:%d\n", file,
3006 + lower_file, bindex, fstart, fend);
3007 + }
3008 + } else { /* lower_file == NULL */
3009 + if (bindex >= fstart && bindex <= fend) {
3010 + /*
3011 + * directories can have NULL lower inodes in
3012 + * b/t start/end, but NOT if at the
3013 + * start/end range.
3014 + */
3015 + if (unlikely(!(S_ISDIR(inode->i_mode) &&
3016 + bindex > fstart &&
3017 + bindex < fend))) {
3018 + PRINT_CALLER(fname, fxn, line);
3019 + pr_debug(" CF6: file/lower=%p:%p "
3020 + "bindex=%d fstart/end=%d:%d\n",
3021 + file, lower_file, bindex,
3022 + fstart, fend);
3023 + }
3024 + }
3025 + }
3026 + }
3027 +
3028 + __unionfs_check_dentry(dentry, fname, fxn, line);
3029 +}
3030 +
3031 +void __unionfs_check_nd(const struct nameidata *nd,
3032 + const char *fname, const char *fxn, int line)
3033 +{
3034 + struct file *file;
3035 + int printed_caller = 0;
3036 +
3037 + if (unlikely(!nd))
3038 + return;
3039 + if (nd->flags & LOOKUP_OPEN) {
3040 + file = nd->intent.open.file;
3041 + if (unlikely(file->f_path.dentry &&
3042 + strcmp(file->f_path.dentry->d_sb->s_type->name,
3043 + UNIONFS_NAME))) {
3044 + PRINT_CALLER(fname, fxn, line);
3045 + pr_debug(" CND1: lower_file of type %s\n",
3046 + file->f_path.dentry->d_sb->s_type->name);
3047 + BUG();
3048 + }
3049 + }
3050 +}
3051 +
3052 +/* useful to track vfsmount leaks that could cause EBUSY on unmount */
3053 +void __show_branch_counts(const struct super_block *sb,
3054 + const char *file, const char *fxn, int line)
3055 +{
3056 + int i;
3057 + struct vfsmount *mnt;
3058 +
3059 + pr_debug("BC:");
3060 + for (i = 0; i < sbmax(sb); i++) {
3061 + if (likely(sb->s_root))
3062 + mnt = UNIONFS_D(sb->s_root)->lower_paths[i].mnt;
3063 + else
3064 + mnt = NULL;
3065 + printk(KERN_CONT "%d:",
3066 + (mnt ? atomic_read(&mnt->mnt_count) : -99));
3067 + }
3068 + printk(KERN_CONT "%s:%s:%d\n", file, fxn, line);
3069 +}
3070 +
3071 +void __show_inode_times(const struct inode *inode,
3072 + const char *file, const char *fxn, int line)
3073 +{
3074 + struct inode *lower_inode;
3075 + int bindex;
3076 +
3077 + for (bindex = ibstart(inode); bindex <= ibend(inode); bindex++) {
3078 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
3079 + if (unlikely(!lower_inode))
3080 + continue;
3081 + pr_debug("IT(%lu:%d): %s:%s:%d "
3082 + "um=%lu/%lu lm=%lu/%lu uc=%lu/%lu lc=%lu/%lu\n",
3083 + inode->i_ino, bindex,
3084 + file, fxn, line,
3085 + inode->i_mtime.tv_sec, inode->i_mtime.tv_nsec,
3086 + lower_inode->i_mtime.tv_sec,
3087 + lower_inode->i_mtime.tv_nsec,
3088 + inode->i_ctime.tv_sec, inode->i_ctime.tv_nsec,
3089 + lower_inode->i_ctime.tv_sec,
3090 + lower_inode->i_ctime.tv_nsec);
3091 + }
3092 +}
3093 +
3094 +void __show_dinode_times(const struct dentry *dentry,
3095 + const char *file, const char *fxn, int line)
3096 +{
3097 + struct inode *inode = dentry->d_inode;
3098 + struct inode *lower_inode;
3099 + int bindex;
3100 +
3101 + for (bindex = ibstart(inode); bindex <= ibend(inode); bindex++) {
3102 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
3103 + if (!lower_inode)
3104 + continue;
3105 + pr_debug("DT(%s:%lu:%d): %s:%s:%d "
3106 + "um=%lu/%lu lm=%lu/%lu uc=%lu/%lu lc=%lu/%lu\n",
3107 + dentry->d_name.name, inode->i_ino, bindex,
3108 + file, fxn, line,
3109 + inode->i_mtime.tv_sec, inode->i_mtime.tv_nsec,
3110 + lower_inode->i_mtime.tv_sec,
3111 + lower_inode->i_mtime.tv_nsec,
3112 + inode->i_ctime.tv_sec, inode->i_ctime.tv_nsec,
3113 + lower_inode->i_ctime.tv_sec,
3114 + lower_inode->i_ctime.tv_nsec);
3115 + }
3116 +}
3117 +
3118 +void __show_inode_counts(const struct inode *inode,
3119 + const char *file, const char *fxn, int line)
3120 +{
3121 + struct inode *lower_inode;
3122 + int bindex;
3123 +
3124 + if (unlikely(!inode)) {
3125 + pr_debug("SiC: Null inode\n");
3126 + return;
3127 + }
3128 + for (bindex = sbstart(inode->i_sb); bindex <= sbend(inode->i_sb);
3129 + bindex++) {
3130 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
3131 + if (unlikely(!lower_inode))
3132 + continue;
3133 + pr_debug("SIC(%lu:%d:%d): lc=%d %s:%s:%d\n",
3134 + inode->i_ino, bindex,
3135 + atomic_read(&(inode)->i_count),
3136 + atomic_read(&(lower_inode)->i_count),
3137 + file, fxn, line);
3138 + }
3139 +}
3140 diff -Naur linux-2.6.30/fs/unionfs/dentry.c linux-2.6.30-magellan/fs/unionfs/dentry.c
3141 --- linux-2.6.30/fs/unionfs/dentry.c 1970-01-01 01:00:00.000000000 +0100
3142 +++ linux-2.6.30-magellan/fs/unionfs/dentry.c 2009-06-11 18:39:19.000000000 +0200
3143 @@ -0,0 +1,397 @@
3144 +/*
3145 + * Copyright (c) 2003-2009 Erez Zadok
3146 + * Copyright (c) 2003-2006 Charles P. Wright
3147 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
3148 + * Copyright (c) 2005-2006 Junjiro Okajima
3149 + * Copyright (c) 2005 Arun M. Krishnakumar
3150 + * Copyright (c) 2004-2006 David P. Quigley
3151 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
3152 + * Copyright (c) 2003 Puja Gupta
3153 + * Copyright (c) 2003 Harikesavan Krishnan
3154 + * Copyright (c) 2003-2009 Stony Brook University
3155 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
3156 + *
3157 + * This program is free software; you can redistribute it and/or modify
3158 + * it under the terms of the GNU General Public License version 2 as
3159 + * published by the Free Software Foundation.
3160 + */
3161 +
3162 +#include "union.h"
3163 +
3164 +bool is_negative_lower(const struct dentry *dentry)
3165 +{
3166 + int bindex;
3167 + struct dentry *lower_dentry;
3168 +
3169 + BUG_ON(!dentry);
3170 + /* cache coherency: check if file was deleted on lower branch */
3171 + if (dbstart(dentry) < 0)
3172 + return true;
3173 + for (bindex = dbstart(dentry); bindex <= dbend(dentry); bindex++) {
3174 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
3175 + /* unhashed (i.e., unlinked) lower dentries don't count */
3176 + if (lower_dentry && lower_dentry->d_inode &&
3177 + !d_deleted(lower_dentry) &&
3178 + !(lower_dentry->d_flags & DCACHE_NFSFS_RENAMED))
3179 + return false;
3180 + }
3181 + return true;
3182 +}
3183 +
3184 +static inline void __dput_lowers(struct dentry *dentry, int start, int end)
3185 +{
3186 + struct dentry *lower_dentry;
3187 + int bindex;
3188 +
3189 + if (start < 0)
3190 + return;
3191 + for (bindex = start; bindex <= end; bindex++) {
3192 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
3193 + if (!lower_dentry)
3194 + continue;
3195 + unionfs_set_lower_dentry_idx(dentry, bindex, NULL);
3196 + dput(lower_dentry);
3197 + }
3198 +}
3199 +
3200 +/*
3201 + * Purge and invalidate as many data pages of a unionfs inode. This is
3202 + * called when the lower inode has changed, and we want to force processes
3203 + * to re-get the new data.
3204 + */
3205 +static inline void purge_inode_data(struct inode *inode)
3206 +{
3207 + /* remove all non-private mappings */
3208 + unmap_mapping_range(inode->i_mapping, 0, 0, 0);
3209 + /* invalidate as many pages as possible */
3210 + invalidate_mapping_pages(inode->i_mapping, 0, -1);
3211 + /*
3212 + * Don't try to truncate_inode_pages here, because this could lead
3213 + * to a deadlock between some of address_space ops and dentry
3214 + * revalidation: the address space op is invoked with a lock on our
3215 + * own page, and truncate_inode_pages will block on locked pages.
3216 + */
3217 +}
3218 +
3219 +/*
3220 + * Revalidate a single file/symlink/special dentry. Assume that info nodes
3221 + * of the @dentry and its @parent are locked. Assume parent is valid,
3222 + * otherwise return false (and let's hope the VFS will try to re-lookup this
3223 + * dentry). Returns true if valid, false otherwise.
3224 + */
3225 +bool __unionfs_d_revalidate(struct dentry *dentry, struct dentry *parent,
3226 + bool willwrite)
3227 +{
3228 + bool valid = true; /* default is valid */
3229 + struct dentry *lower_dentry;
3230 + struct dentry *result;
3231 + int bindex, bstart, bend;
3232 + int sbgen, dgen, pdgen;
3233 + int positive = 0;
3234 + int interpose_flag;
3235 +
3236 + verify_locked(dentry);
3237 + verify_locked(parent);
3238 +
3239 + /* if the dentry is unhashed, do NOT revalidate */
3240 + if (d_deleted(dentry))
3241 + goto out;
3242 +
3243 + dgen = atomic_read(&UNIONFS_D(dentry)->generation);
3244 +
3245 + if (is_newer_lower(dentry)) {
3246 + /* root dentry is always valid */
3247 + if (IS_ROOT(dentry)) {
3248 + unionfs_copy_attr_times(dentry->d_inode);
3249 + } else {
3250 + /*
3251 + * reset generation number to zero, guaranteed to be
3252 + * "old"
3253 + */
3254 + dgen = 0;
3255 + atomic_set(&UNIONFS_D(dentry)->generation, dgen);
3256 + }
3257 + if (!willwrite)
3258 + purge_inode_data(dentry->d_inode);
3259 + }
3260 +
3261 + sbgen = atomic_read(&UNIONFS_SB(dentry->d_sb)->generation);
3262 +
3263 + BUG_ON(dbstart(dentry) == -1);
3264 + if (dentry->d_inode)
3265 + positive = 1;
3266 +
3267 + /* if our dentry is valid, then validate all lower ones */
3268 + if (sbgen == dgen)
3269 + goto validate_lowers;
3270 +
3271 + /* The root entry should always be valid */
3272 + BUG_ON(IS_ROOT(dentry));
3273 +
3274 + /* We can't work correctly if our parent isn't valid. */
3275 + pdgen = atomic_read(&UNIONFS_D(parent)->generation);
3276 +
3277 + /* Free the pointers for our inodes and this dentry. */
3278 + path_put_lowers_all(dentry, false);
3279 +
3280 + interpose_flag = INTERPOSE_REVAL_NEG;
3281 + if (positive) {
3282 + interpose_flag = INTERPOSE_REVAL;
3283 + iput_lowers_all(dentry->d_inode, true);
3284 + }
3285 +
3286 + if (realloc_dentry_private_data(dentry) != 0) {
3287 + valid = false;
3288 + goto out;
3289 + }
3290 +
3291 + result = unionfs_lookup_full(dentry, parent, interpose_flag);
3292 + if (result) {
3293 + if (IS_ERR(result)) {
3294 + valid = false;
3295 + goto out;
3296 + }
3297 + /*
3298 + * current unionfs_lookup_backend() doesn't return
3299 + * a valid dentry
3300 + */
3301 + dput(dentry);
3302 + dentry = result;
3303 + }
3304 +
3305 + if (unlikely(positive && is_negative_lower(dentry))) {
3306 + /* call make_bad_inode here ? */
3307 + d_drop(dentry);
3308 + valid = false;
3309 + goto out;
3310 + }
3311 +
3312 + /*
3313 + * if we got here then we have revalidated our dentry and all lower
3314 + * ones, so we can return safely.
3315 + */
3316 + if (!valid) /* lower dentry revalidation failed */
3317 + goto out;
3318 +
3319 + /*
3320 + * If the parent's gen no. matches the superblock's gen no., then
3321 + * we can update our denty's gen no. If they didn't match, then it
3322 + * was OK to revalidate this dentry with a stale parent, but we'll
3323 + * purposely not update our dentry's gen no. (so it can be redone);
3324 + * and, we'll mark our parent dentry as invalid so it'll force it
3325 + * (and our dentry) to be revalidated.
3326 + */
3327 + if (pdgen == sbgen)
3328 + atomic_set(&UNIONFS_D(dentry)->generation, sbgen);
3329 + goto out;
3330 +
3331 +validate_lowers:
3332 +
3333 + /* The revalidation must occur across all branches */
3334 + bstart = dbstart(dentry);
3335 + bend = dbend(dentry);
3336 + BUG_ON(bstart == -1);
3337 + for (bindex = bstart; bindex <= bend; bindex++) {
3338 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
3339 + if (!lower_dentry || !lower_dentry->d_op
3340 + || !lower_dentry->d_op->d_revalidate)
3341 + continue;
3342 + /*
3343 + * Don't pass nameidata to lower file system, because we
3344 + * don't want an arbitrary lower file being opened or
3345 + * returned to us: it may be useless to us because of the
3346 + * fanout nature of unionfs (cf. file/directory open-file
3347 + * invariants). We will open lower files as and when needed
3348 + * later on.
3349 + */
3350 + if (!lower_dentry->d_op->d_revalidate(lower_dentry, NULL))
3351 + valid = false;
3352 + }
3353 +
3354 + if (!dentry->d_inode ||
3355 + ibstart(dentry->d_inode) < 0 ||
3356 + ibend(dentry->d_inode) < 0) {
3357 + valid = false;
3358 + goto out;
3359 + }
3360 +
3361 + if (valid) {
3362 + /*
3363 + * If we get here, and we copy the meta-data from the lower
3364 + * inode to our inode, then it is vital that we have already
3365 + * purged all unionfs-level file data. We do that in the
3366 + * caller (__unionfs_d_revalidate) by calling
3367 + * purge_inode_data.
3368 + */
3369 + unionfs_copy_attr_all(dentry->d_inode,
3370 + unionfs_lower_inode(dentry->d_inode));
3371 + fsstack_copy_inode_size(dentry->d_inode,
3372 + unionfs_lower_inode(dentry->d_inode));
3373 + }
3374 +
3375 +out:
3376 + return valid;
3377 +}
3378 +
3379 +/*
3380 + * Determine if the lower inode objects have changed from below the unionfs
3381 + * inode. Return true if changed, false otherwise.
3382 + *
3383 + * We check if the mtime or ctime have changed. However, the inode times
3384 + * can be changed by anyone without much protection, including
3385 + * asynchronously. This can sometimes cause unionfs to find that the lower
3386 + * file system doesn't change its inode times quick enough, resulting in a
3387 + * false positive indication (which is harmless, it just makes unionfs do
3388 + * extra work in re-validating the objects). To minimize the chances of
3389 + * these situations, we still consider such small time changes valid, but we
3390 + * don't print debugging messages unless the time changes are greater than
3391 + * UNIONFS_MIN_CC_TIME (which defaults to 3 seconds, as with NFS's acregmin)
3392 + * because significant changes are more likely due to users manually
3393 + * touching lower files.
3394 + */
3395 +bool is_newer_lower(const struct dentry *dentry)
3396 +{
3397 + int bindex;
3398 + struct inode *inode;
3399 + struct inode *lower_inode;
3400 +
3401 + /* ignore if we're called on semi-initialized dentries/inodes */
3402 + if (!dentry || !UNIONFS_D(dentry))
3403 + return false;
3404 + inode = dentry->d_inode;
3405 + if (!inode || !UNIONFS_I(inode)->lower_inodes ||
3406 + ibstart(inode) < 0 || ibend(inode) < 0)
3407 + return false;
3408 +
3409 + for (bindex = ibstart(inode); bindex <= ibend(inode); bindex++) {
3410 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
3411 + if (!lower_inode)
3412 + continue;
3413 +
3414 + /* check if mtime/ctime have changed */
3415 + if (unlikely(timespec_compare(&inode->i_mtime,
3416 + &lower_inode->i_mtime) < 0)) {
3417 + if ((lower_inode->i_mtime.tv_sec -
3418 + inode->i_mtime.tv_sec) > UNIONFS_MIN_CC_TIME) {
3419 + pr_info("unionfs: new lower inode mtime "
3420 + "(bindex=%d, name=%s)\n", bindex,
3421 + dentry->d_name.name);
3422 + show_dinode_times(dentry);
3423 + }
3424 + return true;
3425 + }
3426 + if (unlikely(timespec_compare(&inode->i_ctime,
3427 + &lower_inode->i_ctime) < 0)) {
3428 + if ((lower_inode->i_ctime.tv_sec -
3429 + inode->i_ctime.tv_sec) > UNIONFS_MIN_CC_TIME) {
3430 + pr_info("unionfs: new lower inode ctime "
3431 + "(bindex=%d, name=%s)\n", bindex,
3432 + dentry->d_name.name);
3433 + show_dinode_times(dentry);
3434 + }
3435 + return true;
3436 + }
3437 + }
3438 +
3439 + /*
3440 + * Last check: if this is a positive dentry, but somehow all lower
3441 + * dentries are negative or unhashed, then this dentry needs to be
3442 + * revalidated, because someone probably deleted the objects from
3443 + * the lower branches directly.
3444 + */
3445 + if (is_negative_lower(dentry))
3446 + return true;
3447 +
3448 + return false; /* default: lower is not newer */
3449 +}
3450 +
3451 +static int unionfs_d_revalidate(struct dentry *dentry,
3452 + struct nameidata *nd_unused)
3453 +{
3454 + bool valid = true;
3455 + int err = 1; /* 1 means valid for the VFS */
3456 + struct dentry *parent;
3457 +
3458 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
3459 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
3460 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
3461 +
3462 + valid = __unionfs_d_revalidate(dentry, parent, false);
3463 + if (valid) {
3464 + unionfs_postcopyup_setmnt(dentry);
3465 + unionfs_check_dentry(dentry);
3466 + } else {
3467 + d_drop(dentry);
3468 + err = valid;
3469 + }
3470 + unionfs_unlock_dentry(dentry);
3471 + unionfs_unlock_parent(dentry, parent);
3472 + unionfs_read_unlock(dentry->d_sb);
3473 +
3474 + return err;
3475 +}
3476 +
3477 +static void unionfs_d_release(struct dentry *dentry)
3478 +{
3479 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
3480 + if (unlikely(!UNIONFS_D(dentry)))
3481 + goto out; /* skip if no lower branches */
3482 + /* must lock our branch configuration here */
3483 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
3484 +
3485 + unionfs_check_dentry(dentry);
3486 + /* this could be a negative dentry, so check first */
3487 + if (dbstart(dentry) < 0) {
3488 + unionfs_unlock_dentry(dentry);
3489 + goto out; /* due to a (normal) failed lookup */
3490 + }
3491 +
3492 + /* Release all the lower dentries */
3493 + path_put_lowers_all(dentry, true);
3494 +
3495 + unionfs_unlock_dentry(dentry);
3496 +
3497 +out:
3498 + free_dentry_private_data(dentry);
3499 + unionfs_read_unlock(dentry->d_sb);
3500 + return;
3501 +}
3502 +
3503 +/*
3504 + * Called when we're removing the last reference to our dentry. So we
3505 + * should drop all lower references too.
3506 + */
3507 +static void unionfs_d_iput(struct dentry *dentry, struct inode *inode)
3508 +{
3509 + int rc;
3510 +
3511 + BUG_ON(!dentry);
3512 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
3513 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
3514 +
3515 + if (!UNIONFS_D(dentry) || dbstart(dentry) < 0)
3516 + goto drop_lower_inodes;
3517 + path_put_lowers_all(dentry, false);
3518 +
3519 +drop_lower_inodes:
3520 + rc = atomic_read(&inode->i_count);
3521 + if (rc == 1 && inode->i_nlink == 1 && ibstart(inode) >= 0) {
3522 + /* see Documentation/filesystems/unionfs/issues.txt */
3523 + lockdep_off();
3524 + iput(unionfs_lower_inode(inode));
3525 + lockdep_on();
3526 + unionfs_set_lower_inode(inode, NULL);
3527 + /* XXX: may need to set start/end to -1? */
3528 + }
3529 +
3530 + iput(inode);
3531 +
3532 + unionfs_unlock_dentry(dentry);
3533 + unionfs_read_unlock(dentry->d_sb);
3534 +}
3535 +
3536 +struct dentry_operations unionfs_dops = {
3537 + .d_revalidate = unionfs_d_revalidate,
3538 + .d_release = unionfs_d_release,
3539 + .d_iput = unionfs_d_iput,
3540 +};
3541 diff -Naur linux-2.6.30/fs/unionfs/dirfops.c linux-2.6.30-magellan/fs/unionfs/dirfops.c
3542 --- linux-2.6.30/fs/unionfs/dirfops.c 1970-01-01 01:00:00.000000000 +0100
3543 +++ linux-2.6.30-magellan/fs/unionfs/dirfops.c 2009-06-11 18:39:19.000000000 +0200
3544 @@ -0,0 +1,302 @@
3545 +/*
3546 + * Copyright (c) 2003-2009 Erez Zadok
3547 + * Copyright (c) 2003-2006 Charles P. Wright
3548 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
3549 + * Copyright (c) 2005-2006 Junjiro Okajima
3550 + * Copyright (c) 2005 Arun M. Krishnakumar
3551 + * Copyright (c) 2004-2006 David P. Quigley
3552 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
3553 + * Copyright (c) 2003 Puja Gupta
3554 + * Copyright (c) 2003 Harikesavan Krishnan
3555 + * Copyright (c) 2003-2009 Stony Brook University
3556 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
3557 + *
3558 + * This program is free software; you can redistribute it and/or modify
3559 + * it under the terms of the GNU General Public License version 2 as
3560 + * published by the Free Software Foundation.
3561 + */
3562 +
3563 +#include "union.h"
3564 +
3565 +/* Make sure our rdstate is playing by the rules. */
3566 +static void verify_rdstate_offset(struct unionfs_dir_state *rdstate)
3567 +{
3568 + BUG_ON(rdstate->offset >= DIREOF);
3569 + BUG_ON(rdstate->cookie >= MAXRDCOOKIE);
3570 +}
3571 +
3572 +struct unionfs_getdents_callback {
3573 + struct unionfs_dir_state *rdstate;
3574 + void *dirent;
3575 + int entries_written;
3576 + int filldir_called;
3577 + int filldir_error;
3578 + filldir_t filldir;
3579 + struct super_block *sb;
3580 +};
3581 +
3582 +/* based on generic filldir in fs/readir.c */
3583 +static int unionfs_filldir(void *dirent, const char *oname, int namelen,
3584 + loff_t offset, u64 ino, unsigned int d_type)
3585 +{
3586 + struct unionfs_getdents_callback *buf = dirent;
3587 + struct filldir_node *found = NULL;
3588 + int err = 0;
3589 + int is_whiteout;
3590 + char *name = (char *) oname;
3591 +
3592 + buf->filldir_called++;
3593 +
3594 + is_whiteout = is_whiteout_name(&name, &namelen);
3595 +
3596 + found = find_filldir_node(buf->rdstate, name, namelen, is_whiteout);
3597 +
3598 + if (found) {
3599 + /*
3600 + * If we had non-whiteout entry in dir cache, then mark it
3601 + * as a whiteout and but leave it in the dir cache.
3602 + */
3603 + if (is_whiteout && !found->whiteout)
3604 + found->whiteout = is_whiteout;
3605 + goto out;
3606 + }
3607 +
3608 + /* if 'name' isn't a whiteout, filldir it. */
3609 + if (!is_whiteout) {
3610 + off_t pos = rdstate2offset(buf->rdstate);
3611 + u64 unionfs_ino = ino;
3612 +
3613 + err = buf->filldir(buf->dirent, name, namelen, pos,
3614 + unionfs_ino, d_type);
3615 + buf->rdstate->offset++;
3616 + verify_rdstate_offset(buf->rdstate);
3617 + }
3618 + /*
3619 + * If we did fill it, stuff it in our hash, otherwise return an
3620 + * error.
3621 + */
3622 + if (err) {
3623 + buf->filldir_error = err;
3624 + goto out;
3625 + }
3626 + buf->entries_written++;
3627 + err = add_filldir_node(buf->rdstate, name, namelen,
3628 + buf->rdstate->bindex, is_whiteout);
3629 + if (err)
3630 + buf->filldir_error = err;
3631 +
3632 +out:
3633 + return err;
3634 +}
3635 +
3636 +static int unionfs_readdir(struct file *file, void *dirent, filldir_t filldir)
3637 +{
3638 + int err = 0;
3639 + struct file *lower_file = NULL;
3640 + struct dentry *dentry = file->f_path.dentry;
3641 + struct dentry *parent;
3642 + struct inode *inode = NULL;
3643 + struct unionfs_getdents_callback buf;
3644 + struct unionfs_dir_state *uds;
3645 + int bend;
3646 + loff_t offset;
3647 +
3648 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
3649 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
3650 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
3651 +
3652 + err = unionfs_file_revalidate(file, parent, false);
3653 + if (unlikely(err))
3654 + goto out;
3655 +
3656 + inode = dentry->d_inode;
3657 +
3658 + uds = UNIONFS_F(file)->rdstate;
3659 + if (!uds) {
3660 + if (file->f_pos == DIREOF) {
3661 + goto out;
3662 + } else if (file->f_pos > 0) {
3663 + uds = find_rdstate(inode, file->f_pos);
3664 + if (unlikely(!uds)) {
3665 + err = -ESTALE;
3666 + goto out;
3667 + }
3668 + UNIONFS_F(file)->rdstate = uds;
3669 + } else {
3670 + init_rdstate(file);
3671 + uds = UNIONFS_F(file)->rdstate;
3672 + }
3673 + }
3674 + bend = fbend(file);
3675 +
3676 + while (uds->bindex <= bend) {
3677 + lower_file = unionfs_lower_file_idx(file, uds->bindex);
3678 + if (!lower_file) {
3679 + uds->bindex++;
3680 + uds->dirpos = 0;
3681 + continue;
3682 + }
3683 +
3684 + /* prepare callback buffer */
3685 + buf.filldir_called = 0;
3686 + buf.filldir_error = 0;
3687 + buf.entries_written = 0;
3688 + buf.dirent = dirent;
3689 + buf.filldir = filldir;
3690 + buf.rdstate = uds;
3691 + buf.sb = inode->i_sb;
3692 +
3693 + /* Read starting from where we last left off. */
3694 + offset = vfs_llseek(lower_file, uds->dirpos, SEEK_SET);
3695 + if (offset < 0) {
3696 + err = offset;
3697 + goto out;
3698 + }
3699 + err = vfs_readdir(lower_file, unionfs_filldir, &buf);
3700 +
3701 + /* Save the position for when we continue. */
3702 + offset = vfs_llseek(lower_file, 0, SEEK_CUR);
3703 + if (offset < 0) {
3704 + err = offset;
3705 + goto out;
3706 + }
3707 + uds->dirpos = offset;
3708 +
3709 + /* Copy the atime. */
3710 + fsstack_copy_attr_atime(inode,
3711 + lower_file->f_path.dentry->d_inode);
3712 +
3713 + if (err < 0)
3714 + goto out;
3715 +
3716 + if (buf.filldir_error)
3717 + break;
3718 +
3719 + if (!buf.entries_written) {
3720 + uds->bindex++;
3721 + uds->dirpos = 0;
3722 + }
3723 + }
3724 +
3725 + if (!buf.filldir_error && uds->bindex >= bend) {
3726 + /* Save the number of hash entries for next time. */
3727 + UNIONFS_I(inode)->hashsize = uds->hashentries;
3728 + free_rdstate(uds);
3729 + UNIONFS_F(file)->rdstate = NULL;
3730 + file->f_pos = DIREOF;
3731 + } else {
3732 + file->f_pos = rdstate2offset(uds);
3733 + }
3734 +
3735 +out:
3736 + if (!err)
3737 + unionfs_check_file(file);
3738 + unionfs_unlock_dentry(dentry);
3739 + unionfs_unlock_parent(dentry, parent);
3740 + unionfs_read_unlock(dentry->d_sb);
3741 + return err;
3742 +}
3743 +
3744 +/*
3745 + * This is not meant to be a generic repositioning function. If you do
3746 + * things that aren't supported, then we return EINVAL.
3747 + *
3748 + * What is allowed:
3749 + * (1) seeking to the same position that you are currently at
3750 + * This really has no effect, but returns where you are.
3751 + * (2) seeking to the beginning of the file
3752 + * This throws out all state, and lets you begin again.
3753 + */
3754 +static loff_t unionfs_dir_llseek(struct file *file, loff_t offset, int origin)
3755 +{
3756 + struct unionfs_dir_state *rdstate;
3757 + struct dentry *dentry = file->f_path.dentry;
3758 + struct dentry *parent;
3759 + loff_t err;
3760 +
3761 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
3762 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
3763 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
3764 +
3765 + err = unionfs_file_revalidate(file, parent, false);
3766 + if (unlikely(err))
3767 + goto out;
3768 +
3769 + rdstate = UNIONFS_F(file)->rdstate;
3770 +
3771 + /*
3772 + * we let users seek to their current position, but not anywhere
3773 + * else.
3774 + */
3775 + if (!offset) {
3776 + switch (origin) {
3777 + case SEEK_SET:
3778 + if (rdstate) {
3779 + free_rdstate(rdstate);
3780 + UNIONFS_F(file)->rdstate = NULL;
3781 + }
3782 + init_rdstate(file);
3783 + err = 0;
3784 + break;
3785 + case SEEK_CUR:
3786 + err = file->f_pos;
3787 + break;
3788 + case SEEK_END:
3789 + /* Unsupported, because we would break everything. */
3790 + err = -EINVAL;
3791 + break;
3792 + }
3793 + } else {
3794 + switch (origin) {
3795 + case SEEK_SET:
3796 + if (rdstate) {
3797 + if (offset == rdstate2offset(rdstate))
3798 + err = offset;
3799 + else if (file->f_pos == DIREOF)
3800 + err = DIREOF;
3801 + else
3802 + err = -EINVAL;
3803 + } else {
3804 + struct inode *inode;
3805 + inode = dentry->d_inode;
3806 + rdstate = find_rdstate(inode, offset);
3807 + if (rdstate) {
3808 + UNIONFS_F(file)->rdstate = rdstate;
3809 + err = rdstate->offset;
3810 + } else {
3811 + err = -EINVAL;
3812 + }
3813 + }
3814 + break;
3815 + case SEEK_CUR:
3816 + case SEEK_END:
3817 + /* Unsupported, because we would break everything. */
3818 + err = -EINVAL;
3819 + break;
3820 + }
3821 + }
3822 +
3823 +out:
3824 + if (!err)
3825 + unionfs_check_file(file);
3826 + unionfs_unlock_dentry(dentry);
3827 + unionfs_unlock_parent(dentry, parent);
3828 + unionfs_read_unlock(dentry->d_sb);
3829 + return err;
3830 +}
3831 +
3832 +/*
3833 + * Trimmed directory options, we shouldn't pass everything down since
3834 + * we don't want to operate on partial directories.
3835 + */
3836 +struct file_operations unionfs_dir_fops = {
3837 + .llseek = unionfs_dir_llseek,
3838 + .read = generic_read_dir,
3839 + .readdir = unionfs_readdir,
3840 + .unlocked_ioctl = unionfs_ioctl,
3841 + .open = unionfs_open,
3842 + .release = unionfs_file_release,
3843 + .flush = unionfs_flush,
3844 + .fsync = unionfs_fsync,
3845 + .fasync = unionfs_fasync,
3846 +};
3847 diff -Naur linux-2.6.30/fs/unionfs/dirhelper.c linux-2.6.30-magellan/fs/unionfs/dirhelper.c
3848 --- linux-2.6.30/fs/unionfs/dirhelper.c 1970-01-01 01:00:00.000000000 +0100
3849 +++ linux-2.6.30-magellan/fs/unionfs/dirhelper.c 2009-06-11 18:39:19.000000000 +0200
3850 @@ -0,0 +1,158 @@
3851 +/*
3852 + * Copyright (c) 2003-2009 Erez Zadok
3853 + * Copyright (c) 2003-2006 Charles P. Wright
3854 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
3855 + * Copyright (c) 2005-2006 Junjiro Okajima
3856 + * Copyright (c) 2005 Arun M. Krishnakumar
3857 + * Copyright (c) 2004-2006 David P. Quigley
3858 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
3859 + * Copyright (c) 2003 Puja Gupta
3860 + * Copyright (c) 2003 Harikesavan Krishnan
3861 + * Copyright (c) 2003-2009 Stony Brook University
3862 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
3863 + *
3864 + * This program is free software; you can redistribute it and/or modify
3865 + * it under the terms of the GNU General Public License version 2 as
3866 + * published by the Free Software Foundation.
3867 + */
3868 +
3869 +#include "union.h"
3870 +
3871 +#define RD_NONE 0
3872 +#define RD_CHECK_EMPTY 1
3873 +/* The callback structure for check_empty. */
3874 +struct unionfs_rdutil_callback {
3875 + int err;
3876 + int filldir_called;
3877 + struct unionfs_dir_state *rdstate;
3878 + int mode;
3879 +};
3880 +
3881 +/* This filldir function makes sure only whiteouts exist within a directory. */
3882 +static int readdir_util_callback(void *dirent, const char *oname, int namelen,
3883 + loff_t offset, u64 ino, unsigned int d_type)
3884 +{
3885 + int err = 0;
3886 + struct unionfs_rdutil_callback *buf = dirent;
3887 + int is_whiteout;
3888 + struct filldir_node *found;
3889 + char *name = (char *) oname;
3890 +
3891 + buf->filldir_called = 1;
3892 +
3893 + if (name[0] == '.' && (namelen == 1 ||
3894 + (name[1] == '.' && namelen == 2)))
3895 + goto out;
3896 +
3897 + is_whiteout = is_whiteout_name(&name, &namelen);
3898 +
3899 + found = find_filldir_node(buf->rdstate, name, namelen, is_whiteout);
3900 + /* If it was found in the table there was a previous whiteout. */
3901 + if (found)
3902 + goto out;
3903 +
3904 + /*
3905 + * if it wasn't found and isn't a whiteout, the directory isn't
3906 + * empty.
3907 + */
3908 + err = -ENOTEMPTY;
3909 + if ((buf->mode == RD_CHECK_EMPTY) && !is_whiteout)
3910 + goto out;
3911 +
3912 + err = add_filldir_node(buf->rdstate, name, namelen,
3913 + buf->rdstate->bindex, is_whiteout);
3914 +
3915 +out:
3916 + buf->err = err;
3917 + return err;
3918 +}
3919 +
3920 +/* Is a directory logically empty? */
3921 +int check_empty(struct dentry *dentry, struct dentry *parent,
3922 + struct unionfs_dir_state **namelist)
3923 +{
3924 + int err = 0;
3925 + struct dentry *lower_dentry = NULL;
3926 + struct vfsmount *mnt;
3927 + struct super_block *sb;
3928 + struct file *lower_file;
3929 + struct unionfs_rdutil_callback *buf = NULL;
3930 + int bindex, bstart, bend, bopaque;
3931 +
3932 + sb = dentry->d_sb;
3933 +
3934 +
3935 + BUG_ON(!S_ISDIR(dentry->d_inode->i_mode));
3936 +
3937 + err = unionfs_partial_lookup(dentry, parent);
3938 + if (err)
3939 + goto out;
3940 +
3941 + bstart = dbstart(dentry);
3942 + bend = dbend(dentry);
3943 + bopaque = dbopaque(dentry);
3944 + if (0 <= bopaque && bopaque < bend)
3945 + bend = bopaque;
3946 +
3947 + buf = kmalloc(sizeof(struct unionfs_rdutil_callback), GFP_KERNEL);
3948 + if (unlikely(!buf)) {
3949 + err = -ENOMEM;
3950 + goto out;
3951 + }
3952 + buf->err = 0;
3953 + buf->mode = RD_CHECK_EMPTY;
3954 + buf->rdstate = alloc_rdstate(dentry->d_inode, bstart);
3955 + if (unlikely(!buf->rdstate)) {
3956 + err = -ENOMEM;
3957 + goto out;
3958 + }
3959 +
3960 + /* Process the lower directories with rdutil_callback as a filldir. */
3961 + for (bindex = bstart; bindex <= bend; bindex++) {
3962 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
3963 + if (!lower_dentry)
3964 + continue;
3965 + if (!lower_dentry->d_inode)
3966 + continue;
3967 + if (!S_ISDIR(lower_dentry->d_inode->i_mode))
3968 + continue;
3969 +
3970 + dget(lower_dentry);
3971 + mnt = unionfs_mntget(dentry, bindex);
3972 + branchget(sb, bindex);
3973 + lower_file = dentry_open(lower_dentry, mnt, O_RDONLY, current_cred());
3974 + if (IS_ERR(lower_file)) {
3975 + err = PTR_ERR(lower_file);
3976 + branchput(sb, bindex);
3977 + goto out;
3978 + }
3979 +
3980 + do {
3981 + buf->filldir_called = 0;
3982 + buf->rdstate->bindex = bindex;
3983 + err = vfs_readdir(lower_file,
3984 + readdir_util_callback, buf);
3985 + if (buf->err)
3986 + err = buf->err;
3987 + } while ((err >= 0) && buf->filldir_called);
3988 +
3989 + /* fput calls dput for lower_dentry */
3990 + fput(lower_file);
3991 + branchput(sb, bindex);
3992 +
3993 + if (err < 0)
3994 + goto out;
3995 + }
3996 +
3997 +out:
3998 + if (buf) {
3999 + if (namelist && !err)
4000 + *namelist = buf->rdstate;
4001 + else if (buf->rdstate)
4002 + free_rdstate(buf->rdstate);
4003 + kfree(buf);
4004 + }
4005 +
4006 +
4007 + return err;
4008 +}
4009 diff -Naur linux-2.6.30/fs/unionfs/fanout.h linux-2.6.30-magellan/fs/unionfs/fanout.h
4010 --- linux-2.6.30/fs/unionfs/fanout.h 1970-01-01 01:00:00.000000000 +0100
4011 +++ linux-2.6.30-magellan/fs/unionfs/fanout.h 2009-06-11 18:39:19.000000000 +0200
4012 @@ -0,0 +1,407 @@
4013 +/*
4014 + * Copyright (c) 2003-2009 Erez Zadok
4015 + * Copyright (c) 2003-2006 Charles P. Wright
4016 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
4017 + * Copyright (c) 2005 Arun M. Krishnakumar
4018 + * Copyright (c) 2004-2006 David P. Quigley
4019 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
4020 + * Copyright (c) 2003 Puja Gupta
4021 + * Copyright (c) 2003 Harikesavan Krishnan
4022 + * Copyright (c) 2003-2009 Stony Brook University
4023 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
4024 + *
4025 + * This program is free software; you can redistribute it and/or modify
4026 + * it under the terms of the GNU General Public License version 2 as
4027 + * published by the Free Software Foundation.
4028 + */
4029 +
4030 +#ifndef _FANOUT_H_
4031 +#define _FANOUT_H_
4032 +
4033 +/*
4034 + * Inode to private data
4035 + *
4036 + * Since we use containers and the struct inode is _inside_ the
4037 + * unionfs_inode_info structure, UNIONFS_I will always (given a non-NULL
4038 + * inode pointer), return a valid non-NULL pointer.
4039 + */
4040 +static inline struct unionfs_inode_info *UNIONFS_I(const struct inode *inode)
4041 +{
4042 + return container_of(inode, struct unionfs_inode_info, vfs_inode);
4043 +}
4044 +
4045 +#define ibstart(ino) (UNIONFS_I(ino)->bstart)
4046 +#define ibend(ino) (UNIONFS_I(ino)->bend)
4047 +
4048 +/* Dentry to private data */
4049 +#define UNIONFS_D(dent) ((struct unionfs_dentry_info *)(dent)->d_fsdata)
4050 +#define dbstart(dent) (UNIONFS_D(dent)->bstart)
4051 +#define dbend(dent) (UNIONFS_D(dent)->bend)
4052 +#define dbopaque(dent) (UNIONFS_D(dent)->bopaque)
4053 +
4054 +/* Superblock to private data */
4055 +#define UNIONFS_SB(super) ((struct unionfs_sb_info *)(super)->s_fs_info)
4056 +#define sbstart(sb) 0
4057 +#define sbend(sb) (UNIONFS_SB(sb)->bend)
4058 +#define sbmax(sb) (UNIONFS_SB(sb)->bend + 1)
4059 +#define sbhbid(sb) (UNIONFS_SB(sb)->high_branch_id)
4060 +
4061 +/* File to private Data */
4062 +#define UNIONFS_F(file) ((struct unionfs_file_info *)((file)->private_data))
4063 +#define fbstart(file) (UNIONFS_F(file)->bstart)
4064 +#define fbend(file) (UNIONFS_F(file)->bend)
4065 +
4066 +/* macros to manipulate branch IDs in stored in our superblock */
4067 +static inline int branch_id(struct super_block *sb, int index)
4068 +{
4069 + BUG_ON(!sb || index < 0);
4070 + return UNIONFS_SB(sb)->data[index].branch_id;
4071 +}
4072 +
4073 +static inline void set_branch_id(struct super_block *sb, int index, int val)
4074 +{
4075 + BUG_ON(!sb || index < 0);
4076 + UNIONFS_SB(sb)->data[index].branch_id = val;
4077 +}
4078 +
4079 +static inline void new_branch_id(struct super_block *sb, int index)
4080 +{
4081 + BUG_ON(!sb || index < 0);
4082 + set_branch_id(sb, index, ++UNIONFS_SB(sb)->high_branch_id);
4083 +}
4084 +
4085 +/*
4086 + * Find new index of matching branch with an existing superblock of a known
4087 + * (possibly old) id. This is needed because branches could have been
4088 + * added/deleted causing the branches of any open files to shift.
4089 + *
4090 + * @sb: the new superblock which may have new/different branch IDs
4091 + * @id: the old/existing id we're looking for
4092 + * Returns index of newly found branch (0 or greater), -1 otherwise.
4093 + */
4094 +static inline int branch_id_to_idx(struct super_block *sb, int id)
4095 +{
4096 + int i;
4097 + for (i = 0; i < sbmax(sb); i++) {
4098 + if (branch_id(sb, i) == id)
4099 + return i;
4100 + }
4101 + /* in the non-ODF code, this should really never happen */
4102 + printk(KERN_WARNING "unionfs: cannot find branch with id %d\n", id);
4103 + return -1;
4104 +}
4105 +
4106 +/* File to lower file. */
4107 +static inline struct file *unionfs_lower_file(const struct file *f)
4108 +{
4109 + BUG_ON(!f);
4110 + return UNIONFS_F(f)->lower_files[fbstart(f)];
4111 +}
4112 +
4113 +static inline struct file *unionfs_lower_file_idx(const struct file *f,
4114 + int index)
4115 +{
4116 + BUG_ON(!f || index < 0);
4117 + return UNIONFS_F(f)->lower_files[index];
4118 +}
4119 +
4120 +static inline void unionfs_set_lower_file_idx(struct file *f, int index,
4121 + struct file *val)
4122 +{
4123 + BUG_ON(!f || index < 0);
4124 + UNIONFS_F(f)->lower_files[index] = val;
4125 + /* save branch ID (may be redundant?) */
4126 + UNIONFS_F(f)->saved_branch_ids[index] =
4127 + branch_id((f)->f_path.dentry->d_sb, index);
4128 +}
4129 +
4130 +static inline void unionfs_set_lower_file(struct file *f, struct file *val)
4131 +{
4132 + BUG_ON(!f);
4133 + unionfs_set_lower_file_idx((f), fbstart(f), (val));
4134 +}
4135 +
4136 +/* Inode to lower inode. */
4137 +static inline struct inode *unionfs_lower_inode(const struct inode *i)
4138 +{
4139 + BUG_ON(!i);
4140 + return UNIONFS_I(i)->lower_inodes[ibstart(i)];
4141 +}
4142 +
4143 +static inline struct inode *unionfs_lower_inode_idx(const struct inode *i,
4144 + int index)
4145 +{
4146 + BUG_ON(!i || index < 0);
4147 + return UNIONFS_I(i)->lower_inodes[index];
4148 +}
4149 +
4150 +static inline void unionfs_set_lower_inode_idx(struct inode *i, int index,
4151 + struct inode *val)
4152 +{
4153 + BUG_ON(!i || index < 0);
4154 + UNIONFS_I(i)->lower_inodes[index] = val;
4155 +}
4156 +
4157 +static inline void unionfs_set_lower_inode(struct inode *i, struct inode *val)
4158 +{
4159 + BUG_ON(!i);
4160 + UNIONFS_I(i)->lower_inodes[ibstart(i)] = val;
4161 +}
4162 +
4163 +/* Superblock to lower superblock. */
4164 +static inline struct super_block *unionfs_lower_super(
4165 + const struct super_block *sb)
4166 +{
4167 + BUG_ON(!sb);
4168 + return UNIONFS_SB(sb)->data[sbstart(sb)].sb;
4169 +}
4170 +
4171 +static inline struct super_block *unionfs_lower_super_idx(
4172 + const struct super_block *sb,
4173 + int index)
4174 +{
4175 + BUG_ON(!sb || index < 0);
4176 + return UNIONFS_SB(sb)->data[index].sb;
4177 +}
4178 +
4179 +static inline void unionfs_set_lower_super_idx(struct super_block *sb,
4180 + int index,
4181 + struct super_block *val)
4182 +{
4183 + BUG_ON(!sb || index < 0);
4184 + UNIONFS_SB(sb)->data[index].sb = val;
4185 +}
4186 +
4187 +static inline void unionfs_set_lower_super(struct super_block *sb,
4188 + struct super_block *val)
4189 +{
4190 + BUG_ON(!sb);
4191 + UNIONFS_SB(sb)->data[sbstart(sb)].sb = val;
4192 +}
4193 +
4194 +/* Branch count macros. */
4195 +static inline int branch_count(const struct super_block *sb, int index)
4196 +{
4197 + BUG_ON(!sb || index < 0);
4198 + return atomic_read(&UNIONFS_SB(sb)->data[index].open_files);
4199 +}
4200 +
4201 +static inline void set_branch_count(struct super_block *sb, int index, int val)
4202 +{
4203 + BUG_ON(!sb || index < 0);
4204 + atomic_set(&UNIONFS_SB(sb)->data[index].open_files, val);
4205 +}
4206 +
4207 +static inline void branchget(struct super_block *sb, int index)
4208 +{
4209 + BUG_ON(!sb || index < 0);
4210 + atomic_inc(&UNIONFS_SB(sb)->data[index].open_files);
4211 +}
4212 +
4213 +static inline void branchput(struct super_block *sb, int index)
4214 +{
4215 + BUG_ON(!sb || index < 0);
4216 + atomic_dec(&UNIONFS_SB(sb)->data[index].open_files);
4217 +}
4218 +
4219 +/* Dentry macros */
4220 +static inline void unionfs_set_lower_dentry_idx(struct dentry *dent, int index,
4221 + struct dentry *val)
4222 +{
4223 + BUG_ON(!dent || index < 0);
4224 + UNIONFS_D(dent)->lower_paths[index].dentry = val;
4225 +}
4226 +
4227 +static inline struct dentry *unionfs_lower_dentry_idx(
4228 + const struct dentry *dent,
4229 + int index)
4230 +{
4231 + BUG_ON(!dent || index < 0);
4232 + return UNIONFS_D(dent)->lower_paths[index].dentry;
4233 +}
4234 +
4235 +static inline struct dentry *unionfs_lower_dentry(const struct dentry *dent)
4236 +{
4237 + BUG_ON(!dent);
4238 + return unionfs_lower_dentry_idx(dent, dbstart(dent));
4239 +}
4240 +
4241 +static inline void unionfs_set_lower_mnt_idx(struct dentry *dent, int index,
4242 + struct vfsmount *mnt)
4243 +{
4244 + BUG_ON(!dent || index < 0);
4245 + UNIONFS_D(dent)->lower_paths[index].mnt = mnt;
4246 +}
4247 +
4248 +static inline struct vfsmount *unionfs_lower_mnt_idx(
4249 + const struct dentry *dent,
4250 + int index)
4251 +{
4252 + BUG_ON(!dent || index < 0);
4253 + return UNIONFS_D(dent)->lower_paths[index].mnt;
4254 +}
4255 +
4256 +static inline struct vfsmount *unionfs_lower_mnt(const struct dentry *dent)
4257 +{
4258 + BUG_ON(!dent);
4259 + return unionfs_lower_mnt_idx(dent, dbstart(dent));
4260 +}
4261 +
4262 +/* Macros for locking a dentry. */
4263 +enum unionfs_dentry_lock_class {
4264 + UNIONFS_DMUTEX_NORMAL,
4265 + UNIONFS_DMUTEX_ROOT,
4266 + UNIONFS_DMUTEX_PARENT,
4267 + UNIONFS_DMUTEX_CHILD,
4268 + UNIONFS_DMUTEX_WHITEOUT,
4269 + UNIONFS_DMUTEX_REVAL_PARENT, /* for file/dentry revalidate */
4270 + UNIONFS_DMUTEX_REVAL_CHILD, /* for file/dentry revalidate */
4271 +};
4272 +
4273 +static inline void unionfs_lock_dentry(struct dentry *d,
4274 + unsigned int subclass)
4275 +{
4276 + BUG_ON(!d);
4277 + mutex_lock_nested(&UNIONFS_D(d)->lock, subclass);
4278 +}
4279 +
4280 +static inline void unionfs_unlock_dentry(struct dentry *d)
4281 +{
4282 + BUG_ON(!d);
4283 + mutex_unlock(&UNIONFS_D(d)->lock);
4284 +}
4285 +
4286 +static inline struct dentry *unionfs_lock_parent(struct dentry *d,
4287 + unsigned int subclass)
4288 +{
4289 + struct dentry *p;
4290 +
4291 + BUG_ON(!d);
4292 + p = dget_parent(d);
4293 + if (p != d)
4294 + mutex_lock_nested(&UNIONFS_D(p)->lock, subclass);
4295 + return p;
4296 +}
4297 +
4298 +static inline void unionfs_unlock_parent(struct dentry *d, struct dentry *p)
4299 +{
4300 + BUG_ON(!d);
4301 + BUG_ON(!p);
4302 + if (p != d) {
4303 + BUG_ON(!mutex_is_locked(&UNIONFS_D(p)->lock));
4304 + mutex_unlock(&UNIONFS_D(p)->lock);
4305 + }
4306 + dput(p);
4307 +}
4308 +
4309 +static inline void verify_locked(struct dentry *d)
4310 +{
4311 + BUG_ON(!d);
4312 + BUG_ON(!mutex_is_locked(&UNIONFS_D(d)->lock));
4313 +}
4314 +
4315 +/* macros to put lower objects */
4316 +
4317 +/*
4318 + * iput lower inodes of an unionfs dentry, from bstart to bend. If
4319 + * @free_lower is true, then also kfree the memory used to hold the lower
4320 + * object pointers.
4321 + */
4322 +static inline void iput_lowers(struct inode *inode,
4323 + int bstart, int bend, bool free_lower)
4324 +{
4325 + struct inode *lower_inode;
4326 + int bindex;
4327 +
4328 + BUG_ON(!inode);
4329 + BUG_ON(!UNIONFS_I(inode));
4330 + BUG_ON(bstart < 0);
4331 +
4332 + for (bindex = bstart; bindex <= bend; bindex++) {
4333 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
4334 + if (lower_inode) {
4335 + unionfs_set_lower_inode_idx(inode, bindex, NULL);
4336 + /* see Documentation/filesystems/unionfs/issues.txt */
4337 + lockdep_off();
4338 + iput(lower_inode);
4339 + lockdep_on();
4340 + }
4341 + }
4342 +
4343 + if (free_lower) {
4344 + kfree(UNIONFS_I(inode)->lower_inodes);
4345 + UNIONFS_I(inode)->lower_inodes = NULL;
4346 + }
4347 +}
4348 +
4349 +/* iput all lower inodes, and reset start/end branch indices to -1 */
4350 +static inline void iput_lowers_all(struct inode *inode, bool free_lower)
4351 +{
4352 + int bstart, bend;
4353 +
4354 + BUG_ON(!inode);
4355 + BUG_ON(!UNIONFS_I(inode));
4356 + bstart = ibstart(inode);
4357 + bend = ibend(inode);
4358 + BUG_ON(bstart < 0);
4359 +
4360 + iput_lowers(inode, bstart, bend, free_lower);
4361 + ibstart(inode) = ibend(inode) = -1;
4362 +}
4363 +
4364 +/*
4365 + * dput/mntput all lower dentries and vfsmounts of an unionfs dentry, from
4366 + * bstart to bend. If @free_lower is true, then also kfree the memory used
4367 + * to hold the lower object pointers.
4368 + *
4369 + * XXX: implement using path_put VFS macros
4370 + */
4371 +static inline void path_put_lowers(struct dentry *dentry,
4372 + int bstart, int bend, bool free_lower)
4373 +{
4374 + struct dentry *lower_dentry;
4375 + struct vfsmount *lower_mnt;
4376 + int bindex;
4377 +
4378 + BUG_ON(!dentry);
4379 + BUG_ON(!UNIONFS_D(dentry));
4380 + BUG_ON(bstart < 0);
4381 +
4382 + for (bindex = bstart; bindex <= bend; bindex++) {
4383 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
4384 + if (lower_dentry) {
4385 + unionfs_set_lower_dentry_idx(dentry, bindex, NULL);
4386 + dput(lower_dentry);
4387 + }
4388 + lower_mnt = unionfs_lower_mnt_idx(dentry, bindex);
4389 + if (lower_mnt) {
4390 + unionfs_set_lower_mnt_idx(dentry, bindex, NULL);
4391 + mntput(lower_mnt);
4392 + }
4393 + }
4394 +
4395 + if (free_lower) {
4396 + kfree(UNIONFS_D(dentry)->lower_paths);
4397 + UNIONFS_D(dentry)->lower_paths = NULL;
4398 + }
4399 +}
4400 +
4401 +/*
4402 + * dput/mntput all lower dentries and vfsmounts, and reset start/end branch
4403 + * indices to -1.
4404 + */
4405 +static inline void path_put_lowers_all(struct dentry *dentry, bool free_lower)
4406 +{
4407 + int bstart, bend;
4408 +
4409 + BUG_ON(!dentry);
4410 + BUG_ON(!UNIONFS_D(dentry));
4411 + bstart = dbstart(dentry);
4412 + bend = dbend(dentry);
4413 + BUG_ON(bstart < 0);
4414 +
4415 + path_put_lowers(dentry, bstart, bend, free_lower);
4416 + dbstart(dentry) = dbend(dentry) = -1;
4417 +}
4418 +
4419 +#endif /* not _FANOUT_H */
4420 diff -Naur linux-2.6.30/fs/unionfs/file.c linux-2.6.30-magellan/fs/unionfs/file.c
4421 --- linux-2.6.30/fs/unionfs/file.c 1970-01-01 01:00:00.000000000 +0100
4422 +++ linux-2.6.30-magellan/fs/unionfs/file.c 2009-06-11 18:39:19.000000000 +0200
4423 @@ -0,0 +1,380 @@
4424 +/*
4425 + * Copyright (c) 2003-2009 Erez Zadok
4426 + * Copyright (c) 2003-2006 Charles P. Wright
4427 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
4428 + * Copyright (c) 2005-2006 Junjiro Okajima
4429 + * Copyright (c) 2005 Arun M. Krishnakumar
4430 + * Copyright (c) 2004-2006 David P. Quigley
4431 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
4432 + * Copyright (c) 2003 Puja Gupta
4433 + * Copyright (c) 2003 Harikesavan Krishnan
4434 + * Copyright (c) 2003-2009 Stony Brook University
4435 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
4436 + *
4437 + * This program is free software; you can redistribute it and/or modify
4438 + * it under the terms of the GNU General Public License version 2 as
4439 + * published by the Free Software Foundation.
4440 + */
4441 +
4442 +#include "union.h"
4443 +
4444 +static ssize_t unionfs_read(struct file *file, char __user *buf,
4445 + size_t count, loff_t *ppos)
4446 +{
4447 + int err;
4448 + struct file *lower_file;
4449 + struct dentry *dentry = file->f_path.dentry;
4450 + struct dentry *parent;
4451 +
4452 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
4453 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4454 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4455 +
4456 + err = unionfs_file_revalidate(file, parent, false);
4457 + if (unlikely(err))
4458 + goto out;
4459 +
4460 + lower_file = unionfs_lower_file(file);
4461 + err = vfs_read(lower_file, buf, count, ppos);
4462 + /* update our inode atime upon a successful lower read */
4463 + if (err >= 0) {
4464 + fsstack_copy_attr_atime(dentry->d_inode,
4465 + lower_file->f_path.dentry->d_inode);
4466 + unionfs_check_file(file);
4467 + }
4468 +
4469 +out:
4470 + unionfs_unlock_dentry(dentry);
4471 + unionfs_unlock_parent(dentry, parent);
4472 + unionfs_read_unlock(dentry->d_sb);
4473 + return err;
4474 +}
4475 +
4476 +static ssize_t unionfs_write(struct file *file, const char __user *buf,
4477 + size_t count, loff_t *ppos)
4478 +{
4479 + int err = 0;
4480 + struct file *lower_file;
4481 + struct dentry *dentry = file->f_path.dentry;
4482 + struct dentry *parent;
4483 +
4484 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
4485 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4486 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4487 +
4488 + err = unionfs_file_revalidate(file, parent, true);
4489 + if (unlikely(err))
4490 + goto out;
4491 +
4492 + lower_file = unionfs_lower_file(file);
4493 + err = vfs_write(lower_file, buf, count, ppos);
4494 + /* update our inode times+sizes upon a successful lower write */
4495 + if (err >= 0) {
4496 + fsstack_copy_inode_size(dentry->d_inode,
4497 + lower_file->f_path.dentry->d_inode);
4498 + fsstack_copy_attr_times(dentry->d_inode,
4499 + lower_file->f_path.dentry->d_inode);
4500 + UNIONFS_F(file)->wrote_to_file = true; /* for delayed copyup */
4501 + unionfs_check_file(file);
4502 + }
4503 +
4504 +out:
4505 + unionfs_unlock_dentry(dentry);
4506 + unionfs_unlock_parent(dentry, parent);
4507 + unionfs_read_unlock(dentry->d_sb);
4508 + return err;
4509 +}
4510 +
4511 +static int unionfs_file_readdir(struct file *file, void *dirent,
4512 + filldir_t filldir)
4513 +{
4514 + return -ENOTDIR;
4515 +}
4516 +
4517 +static int unionfs_mmap(struct file *file, struct vm_area_struct *vma)
4518 +{
4519 + int err = 0;
4520 + bool willwrite;
4521 + struct file *lower_file;
4522 + struct dentry *dentry = file->f_path.dentry;
4523 + struct dentry *parent;
4524 + struct vm_operations_struct *saved_vm_ops = NULL;
4525 +
4526 + /*
4527 + * Since mm/memory.c:might_fault() (under PROVE_LOCKING) was
4528 + * modified in 2.6.29-rc1 to call might_lock_read on mmap_sem, this
4529 + * has been causing false positives in file system stacking layers.
4530 + * In particular, our ->mmap is called after sys_mmap2 already holds
4531 + * mmap_sem, then we lock our own mutexes; but earlier, it's
4532 + * possible for lockdep to have locked our mutexes first, and then
4533 + * we call a lower ->readdir which could call might_fault. The
4534 + * different ordering of the locks is what lockdep complains about
4535 + * -- unnecessarily. Therefore, we have no choice but to tell
4536 + * lockdep to temporarily turn off lockdep here. Note: the comments
4537 + * inside might_sleep also suggest that it would have been
4538 + * nicer to only annotate paths that needs that might_lock_read.
4539 + */
4540 + lockdep_off();
4541 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
4542 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4543 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4544 +
4545 + /* This might be deferred to mmap's writepage */
4546 + willwrite = ((vma->vm_flags | VM_SHARED | VM_WRITE) == vma->vm_flags);
4547 + err = unionfs_file_revalidate(file, parent, willwrite);
4548 + if (unlikely(err))
4549 + goto out;
4550 + unionfs_check_file(file);
4551 +
4552 + /*
4553 + * File systems which do not implement ->writepage may use
4554 + * generic_file_readonly_mmap as their ->mmap op. If you call
4555 + * generic_file_readonly_mmap with VM_WRITE, you'd get an -EINVAL.
4556 + * But we cannot call the lower ->mmap op, so we can't tell that
4557 + * writeable mappings won't work. Therefore, our only choice is to
4558 + * check if the lower file system supports the ->writepage, and if
4559 + * not, return EINVAL (the same error that
4560 + * generic_file_readonly_mmap returns in that case).
4561 + */
4562 + lower_file = unionfs_lower_file(file);
4563 + if (willwrite && !lower_file->f_mapping->a_ops->writepage) {
4564 + err = -EINVAL;
4565 + printk(KERN_ERR "unionfs: branch %d file system does not "
4566 + "support writeable mmap\n", fbstart(file));
4567 + goto out;
4568 + }
4569 +
4570 + /*
4571 + * find and save lower vm_ops.
4572 + *
4573 + * XXX: the VFS should have a cleaner way of finding the lower vm_ops
4574 + */
4575 + if (!UNIONFS_F(file)->lower_vm_ops) {
4576 + err = lower_file->f_op->mmap(lower_file, vma);
4577 + if (err) {
4578 + printk(KERN_ERR "unionfs: lower mmap failed %d\n", err);
4579 + goto out;
4580 + }
4581 + saved_vm_ops = vma->vm_ops;
4582 + err = do_munmap(current->mm, vma->vm_start,
4583 + vma->vm_end - vma->vm_start);
4584 + if (err) {
4585 + printk(KERN_ERR "unionfs: do_munmap failed %d\n", err);
4586 + goto out;
4587 + }
4588 + }
4589 +
4590 + file->f_mapping->a_ops = &unionfs_dummy_aops;
4591 + err = generic_file_mmap(file, vma);
4592 + file->f_mapping->a_ops = &unionfs_aops;
4593 + if (err) {
4594 + printk(KERN_ERR "unionfs: generic_file_mmap failed %d\n", err);
4595 + goto out;
4596 + }
4597 + vma->vm_ops = &unionfs_vm_ops;
4598 + if (!UNIONFS_F(file)->lower_vm_ops)
4599 + UNIONFS_F(file)->lower_vm_ops = saved_vm_ops;
4600 +
4601 +out:
4602 + if (!err) {
4603 + /* copyup could cause parent dir times to change */
4604 + unionfs_copy_attr_times(parent->d_inode);
4605 + unionfs_check_file(file);
4606 + }
4607 + unionfs_unlock_dentry(dentry);
4608 + unionfs_unlock_parent(dentry, parent);
4609 + unionfs_read_unlock(dentry->d_sb);
4610 + lockdep_on();
4611 + return err;
4612 +}
4613 +
4614 +int unionfs_fsync(struct file *file, struct dentry *dentry, int datasync)
4615 +{
4616 + int bindex, bstart, bend;
4617 + struct file *lower_file;
4618 + struct dentry *lower_dentry;
4619 + struct dentry *parent;
4620 + struct inode *lower_inode, *inode;
4621 + int err = -EINVAL;
4622 +
4623 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
4624 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4625 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4626 +
4627 + err = unionfs_file_revalidate(file, parent, true);
4628 + if (unlikely(err))
4629 + goto out;
4630 + unionfs_check_file(file);
4631 +
4632 + bstart = fbstart(file);
4633 + bend = fbend(file);
4634 + if (bstart < 0 || bend < 0)
4635 + goto out;
4636 +
4637 + inode = dentry->d_inode;
4638 + if (unlikely(!inode)) {
4639 + printk(KERN_ERR
4640 + "unionfs: null lower inode in unionfs_fsync\n");
4641 + goto out;
4642 + }
4643 + for (bindex = bstart; bindex <= bend; bindex++) {
4644 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
4645 + if (!lower_inode || !lower_inode->i_fop->fsync)
4646 + continue;
4647 + lower_file = unionfs_lower_file_idx(file, bindex);
4648 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
4649 + mutex_lock(&lower_inode->i_mutex);
4650 + err = lower_inode->i_fop->fsync(lower_file,
4651 + lower_dentry,
4652 + datasync);
4653 + if (!err && bindex == bstart)
4654 + fsstack_copy_attr_times(inode, lower_inode);
4655 + mutex_unlock(&lower_inode->i_mutex);
4656 + if (err)
4657 + goto out;
4658 + }
4659 +
4660 +out:
4661 + if (!err)
4662 + unionfs_check_file(file);
4663 + unionfs_unlock_dentry(dentry);
4664 + unionfs_unlock_parent(dentry, parent);
4665 + unionfs_read_unlock(dentry->d_sb);
4666 + return err;
4667 +}
4668 +
4669 +int unionfs_fasync(int fd, struct file *file, int flag)
4670 +{
4671 + int bindex, bstart, bend;
4672 + struct file *lower_file;
4673 + struct dentry *dentry = file->f_path.dentry;
4674 + struct dentry *parent;
4675 + struct inode *lower_inode, *inode;
4676 + int err = 0;
4677 +
4678 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
4679 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4680 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4681 +
4682 + err = unionfs_file_revalidate(file, parent, true);
4683 + if (unlikely(err))
4684 + goto out;
4685 + unionfs_check_file(file);
4686 +
4687 + bstart = fbstart(file);
4688 + bend = fbend(file);
4689 + if (bstart < 0 || bend < 0)
4690 + goto out;
4691 +
4692 + inode = dentry->d_inode;
4693 + if (unlikely(!inode)) {
4694 + printk(KERN_ERR
4695 + "unionfs: null lower inode in unionfs_fasync\n");
4696 + goto out;
4697 + }
4698 + for (bindex = bstart; bindex <= bend; bindex++) {
4699 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
4700 + if (!lower_inode || !lower_inode->i_fop->fasync)
4701 + continue;
4702 + lower_file = unionfs_lower_file_idx(file, bindex);
4703 + mutex_lock(&lower_inode->i_mutex);
4704 + err = lower_inode->i_fop->fasync(fd, lower_file, flag);
4705 + if (!err && bindex == bstart)
4706 + fsstack_copy_attr_times(inode, lower_inode);
4707 + mutex_unlock(&lower_inode->i_mutex);
4708 + if (err)
4709 + goto out;
4710 + }
4711 +
4712 +out:
4713 + if (!err)
4714 + unionfs_check_file(file);
4715 + unionfs_unlock_dentry(dentry);
4716 + unionfs_unlock_parent(dentry, parent);
4717 + unionfs_read_unlock(dentry->d_sb);
4718 + return err;
4719 +}
4720 +
4721 +static ssize_t unionfs_splice_read(struct file *file, loff_t *ppos,
4722 + struct pipe_inode_info *pipe, size_t len,
4723 + unsigned int flags)
4724 +{
4725 + ssize_t err;
4726 + struct file *lower_file;
4727 + struct dentry *dentry = file->f_path.dentry;
4728 + struct dentry *parent;
4729 +
4730 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
4731 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4732 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4733 +
4734 + err = unionfs_file_revalidate(file, parent, false);
4735 + if (unlikely(err))
4736 + goto out;
4737 +
4738 + lower_file = unionfs_lower_file(file);
4739 + err = vfs_splice_to(lower_file, ppos, pipe, len, flags);
4740 + /* update our inode atime upon a successful lower splice-read */
4741 + if (err >= 0) {
4742 + fsstack_copy_attr_atime(dentry->d_inode,
4743 + lower_file->f_path.dentry->d_inode);
4744 + unionfs_check_file(file);
4745 + }
4746 +
4747 +out:
4748 + unionfs_unlock_dentry(dentry);
4749 + unionfs_unlock_parent(dentry, parent);
4750 + unionfs_read_unlock(dentry->d_sb);
4751 + return err;
4752 +}
4753 +
4754 +static ssize_t unionfs_splice_write(struct pipe_inode_info *pipe,
4755 + struct file *file, loff_t *ppos,
4756 + size_t len, unsigned int flags)
4757 +{
4758 + ssize_t err = 0;
4759 + struct file *lower_file;
4760 + struct dentry *dentry = file->f_path.dentry;
4761 + struct dentry *parent;
4762 +
4763 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_PARENT);
4764 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4765 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4766 +
4767 + err = unionfs_file_revalidate(file, parent, true);
4768 + if (unlikely(err))
4769 + goto out;
4770 +
4771 + lower_file = unionfs_lower_file(file);
4772 + err = vfs_splice_from(pipe, lower_file, ppos, len, flags);
4773 + /* update our inode times+sizes upon a successful lower write */
4774 + if (err >= 0) {
4775 + fsstack_copy_inode_size(dentry->d_inode,
4776 + lower_file->f_path.dentry->d_inode);
4777 + fsstack_copy_attr_times(dentry->d_inode,
4778 + lower_file->f_path.dentry->d_inode);
4779 + unionfs_check_file(file);
4780 + }
4781 +
4782 +out:
4783 + unionfs_unlock_dentry(dentry);
4784 + unionfs_unlock_parent(dentry, parent);
4785 + unionfs_read_unlock(dentry->d_sb);
4786 + return err;
4787 +}
4788 +
4789 +struct file_operations unionfs_main_fops = {
4790 + .llseek = generic_file_llseek,
4791 + .read = unionfs_read,
4792 + .write = unionfs_write,
4793 + .readdir = unionfs_file_readdir,
4794 + .unlocked_ioctl = unionfs_ioctl,
4795 + .mmap = unionfs_mmap,
4796 + .open = unionfs_open,
4797 + .flush = unionfs_flush,
4798 + .release = unionfs_file_release,
4799 + .fsync = unionfs_fsync,
4800 + .fasync = unionfs_fasync,
4801 + .splice_read = unionfs_splice_read,
4802 + .splice_write = unionfs_splice_write,
4803 +};
4804 diff -Naur linux-2.6.30/fs/unionfs/inode.c linux-2.6.30-magellan/fs/unionfs/inode.c
4805 --- linux-2.6.30/fs/unionfs/inode.c 1970-01-01 01:00:00.000000000 +0100
4806 +++ linux-2.6.30-magellan/fs/unionfs/inode.c 2009-06-11 18:39:19.000000000 +0200
4807 @@ -0,0 +1,1035 @@
4808 +/*
4809 + * Copyright (c) 2003-2009 Erez Zadok
4810 + * Copyright (c) 2003-2006 Charles P. Wright
4811 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
4812 + * Copyright (c) 2005-2006 Junjiro Okajima
4813 + * Copyright (c) 2005 Arun M. Krishnakumar
4814 + * Copyright (c) 2004-2006 David P. Quigley
4815 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
4816 + * Copyright (c) 2003 Puja Gupta
4817 + * Copyright (c) 2003 Harikesavan Krishnan
4818 + * Copyright (c) 2003-2009 Stony Brook University
4819 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
4820 + *
4821 + * This program is free software; you can redistribute it and/or modify
4822 + * it under the terms of the GNU General Public License version 2 as
4823 + * published by the Free Software Foundation.
4824 + */
4825 +
4826 +#include "union.h"
4827 +
4828 +/*
4829 + * Find a writeable branch to create new object in. Checks all writeble
4830 + * branches of the parent inode, from istart to iend order; if none are
4831 + * suitable, also tries branch 0 (which may require a copyup).
4832 + *
4833 + * Return a lower_dentry we can use to create object in, or ERR_PTR.
4834 + */
4835 +static struct dentry *find_writeable_branch(struct inode *parent,
4836 + struct dentry *dentry)
4837 +{
4838 + int err = -EINVAL;
4839 + int bindex, istart, iend;
4840 + struct dentry *lower_dentry = NULL;
4841 +
4842 + istart = ibstart(parent);
4843 + iend = ibend(parent);
4844 + if (istart < 0)
4845 + goto out;
4846 +
4847 +begin:
4848 + for (bindex = istart; bindex <= iend; bindex++) {
4849 + /* skip non-writeable branches */
4850 + err = is_robranch_super(dentry->d_sb, bindex);
4851 + if (err) {
4852 + err = -EROFS;
4853 + continue;
4854 + }
4855 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
4856 + if (!lower_dentry)
4857 + continue;
4858 + /*
4859 + * check for whiteouts in writeable branch, and remove them
4860 + * if necessary.
4861 + */
4862 + err = check_unlink_whiteout(dentry, lower_dentry, bindex);
4863 + if (err > 0) /* ignore if whiteout found and removed */
4864 + err = 0;
4865 + if (err)
4866 + continue;
4867 + /* if get here, we can write to the branch */
4868 + break;
4869 + }
4870 + /*
4871 + * If istart wasn't already branch 0, and we got any error, then try
4872 + * branch 0 (which may require copyup)
4873 + */
4874 + if (err && istart > 0) {
4875 + istart = iend = 0;
4876 + goto begin;
4877 + }
4878 +
4879 + /*
4880 + * If we tried even branch 0, and still got an error, abort. But if
4881 + * the error was an EROFS, then we should try to copyup.
4882 + */
4883 + if (err && err != -EROFS)
4884 + goto out;
4885 +
4886 + /*
4887 + * If we get here, then check if copyup needed. If lower_dentry is
4888 + * NULL, create the entire dentry directory structure in branch 0.
4889 + */
4890 + if (!lower_dentry) {
4891 + bindex = 0;
4892 + lower_dentry = create_parents(parent, dentry,
4893 + dentry->d_name.name, bindex);
4894 + if (IS_ERR(lower_dentry)) {
4895 + err = PTR_ERR(lower_dentry);
4896 + goto out;
4897 + }
4898 + }
4899 + err = 0; /* all's well */
4900 +out:
4901 + if (err)
4902 + return ERR_PTR(err);
4903 + return lower_dentry;
4904 +}
4905 +
4906 +static int unionfs_create(struct inode *dir, struct dentry *dentry,
4907 + int mode, struct nameidata *nd_unused)
4908 +{
4909 + int err = 0;
4910 + struct dentry *lower_dentry = NULL;
4911 + struct dentry *lower_parent_dentry = NULL;
4912 + struct dentry *parent;
4913 + int valid = 0;
4914 + struct nameidata lower_nd;
4915 +
4916 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
4917 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4918 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
4919 +
4920 + valid = __unionfs_d_revalidate(dentry, parent, false);
4921 + if (unlikely(!valid)) {
4922 + err = -ESTALE; /* same as what real_lookup does */
4923 + goto out;
4924 + }
4925 +
4926 + lower_dentry = find_writeable_branch(dir, dentry);
4927 + if (IS_ERR(lower_dentry)) {
4928 + err = PTR_ERR(lower_dentry);
4929 + goto out;
4930 + }
4931 +
4932 + lower_parent_dentry = lock_parent(lower_dentry);
4933 + if (IS_ERR(lower_parent_dentry)) {
4934 + err = PTR_ERR(lower_parent_dentry);
4935 + goto out;
4936 + }
4937 +
4938 + err = init_lower_nd(&lower_nd, LOOKUP_CREATE);
4939 + if (unlikely(err < 0))
4940 + goto out;
4941 + err = vfs_create(lower_parent_dentry->d_inode, lower_dentry, mode,
4942 + &lower_nd);
4943 + release_lower_nd(&lower_nd, err);
4944 +
4945 + if (!err) {
4946 + err = PTR_ERR(unionfs_interpose(dentry, dir->i_sb, 0));
4947 + if (!err) {
4948 + unionfs_copy_attr_times(dir);
4949 + fsstack_copy_inode_size(dir,
4950 + lower_parent_dentry->d_inode);
4951 + /* update no. of links on parent directory */
4952 + dir->i_nlink = unionfs_get_nlinks(dir);
4953 + }
4954 + }
4955 +
4956 + unlock_dir(lower_parent_dentry);
4957 +
4958 +out:
4959 + if (!err) {
4960 + unionfs_postcopyup_setmnt(dentry);
4961 + unionfs_check_inode(dir);
4962 + unionfs_check_dentry(dentry);
4963 + }
4964 + unionfs_unlock_dentry(dentry);
4965 + unionfs_unlock_parent(dentry, parent);
4966 + unionfs_read_unlock(dentry->d_sb);
4967 + return err;
4968 +}
4969 +
4970 +/*
4971 + * unionfs_lookup is the only special function which takes a dentry, yet we
4972 + * do NOT want to call __unionfs_d_revalidate_chain because by definition,
4973 + * we don't have a valid dentry here yet.
4974 + */
4975 +static struct dentry *unionfs_lookup(struct inode *dir,
4976 + struct dentry *dentry,
4977 + struct nameidata *nd_unused)
4978 +{
4979 + struct dentry *ret, *parent;
4980 + int err = 0;
4981 +
4982 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
4983 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
4984 +
4985 + /*
4986 + * As long as we lock/dget the parent, then can skip validating the
4987 + * parent now; we may have to rebuild this dentry on the next
4988 + * ->d_revalidate, however.
4989 + */
4990 +
4991 + /* allocate dentry private data. We free it in ->d_release */
4992 + err = new_dentry_private_data(dentry, UNIONFS_DMUTEX_CHILD);
4993 + if (unlikely(err)) {
4994 + ret = ERR_PTR(err);
4995 + goto out;
4996 + }
4997 +
4998 + ret = unionfs_lookup_full(dentry, parent, INTERPOSE_LOOKUP);
4999 +
5000 + if (!IS_ERR(ret)) {
5001 + if (ret)
5002 + dentry = ret;
5003 + /* lookup_full can return multiple positive dentries */
5004 + if (dentry->d_inode && !S_ISDIR(dentry->d_inode->i_mode)) {
5005 + BUG_ON(dbstart(dentry) < 0);
5006 + unionfs_postcopyup_release(dentry);
5007 + }
5008 + unionfs_copy_attr_times(dentry->d_inode);
5009 + }
5010 +
5011 + unionfs_check_inode(dir);
5012 + if (!IS_ERR(ret))
5013 + unionfs_check_dentry(dentry);
5014 + unionfs_check_dentry(parent);
5015 + unionfs_unlock_dentry(dentry); /* locked in new_dentry_private data */
5016 +
5017 +out:
5018 + unionfs_unlock_parent(dentry, parent);
5019 + unionfs_read_unlock(dentry->d_sb);
5020 +
5021 + return ret;
5022 +}
5023 +
5024 +static int unionfs_link(struct dentry *old_dentry, struct inode *dir,
5025 + struct dentry *new_dentry)
5026 +{
5027 + int err = 0;
5028 + struct dentry *lower_old_dentry = NULL;
5029 + struct dentry *lower_new_dentry = NULL;
5030 + struct dentry *lower_dir_dentry = NULL;
5031 + struct dentry *old_parent, *new_parent;
5032 + char *name = NULL;
5033 + bool valid;
5034 +
5035 + unionfs_read_lock(old_dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5036 + old_parent = dget_parent(old_dentry);
5037 + new_parent = dget_parent(new_dentry);
5038 + unionfs_double_lock_parents(old_parent, new_parent);
5039 + unionfs_double_lock_dentry(old_dentry, new_dentry);
5040 +
5041 + valid = __unionfs_d_revalidate(old_dentry, old_parent, false);
5042 + if (unlikely(!valid)) {
5043 + err = -ESTALE;
5044 + goto out;
5045 + }
5046 + if (new_dentry->d_inode) {
5047 + valid = __unionfs_d_revalidate(new_dentry, new_parent, false);
5048 + if (unlikely(!valid)) {
5049 + err = -ESTALE;
5050 + goto out;
5051 + }
5052 + }
5053 +
5054 + lower_new_dentry = unionfs_lower_dentry(new_dentry);
5055 +
5056 + /* check for a whiteout in new dentry branch, and delete it */
5057 + err = check_unlink_whiteout(new_dentry, lower_new_dentry,
5058 + dbstart(new_dentry));
5059 + if (err > 0) { /* whiteout found and removed successfully */
5060 + lower_dir_dentry = dget_parent(lower_new_dentry);
5061 + fsstack_copy_attr_times(dir, lower_dir_dentry->d_inode);
5062 + dput(lower_dir_dentry);
5063 + dir->i_nlink = unionfs_get_nlinks(dir);
5064 + err = 0;
5065 + }
5066 + if (err)
5067 + goto out;
5068 +
5069 + /* check if parent hierachy is needed, then link in same branch */
5070 + if (dbstart(old_dentry) != dbstart(new_dentry)) {
5071 + lower_new_dentry = create_parents(dir, new_dentry,
5072 + new_dentry->d_name.name,
5073 + dbstart(old_dentry));
5074 + err = PTR_ERR(lower_new_dentry);
5075 + if (IS_COPYUP_ERR(err))
5076 + goto docopyup;
5077 + if (!lower_new_dentry || IS_ERR(lower_new_dentry))
5078 + goto out;
5079 + }
5080 + lower_new_dentry = unionfs_lower_dentry(new_dentry);
5081 + lower_old_dentry = unionfs_lower_dentry(old_dentry);
5082 +
5083 + BUG_ON(dbstart(old_dentry) != dbstart(new_dentry));
5084 + lower_dir_dentry = lock_parent(lower_new_dentry);
5085 + err = is_robranch(old_dentry);
5086 + if (!err) {
5087 + /* see Documentation/filesystems/unionfs/issues.txt */
5088 + lockdep_off();
5089 + err = vfs_link(lower_old_dentry, lower_dir_dentry->d_inode,
5090 + lower_new_dentry);
5091 + lockdep_on();
5092 + }
5093 + unlock_dir(lower_dir_dentry);
5094 +
5095 +docopyup:
5096 + if (IS_COPYUP_ERR(err)) {
5097 + int old_bstart = dbstart(old_dentry);
5098 + int bindex;
5099 +
5100 + for (bindex = old_bstart - 1; bindex >= 0; bindex--) {
5101 + err = copyup_dentry(old_parent->d_inode,
5102 + old_dentry, old_bstart,
5103 + bindex, old_dentry->d_name.name,
5104 + old_dentry->d_name.len, NULL,
5105 + i_size_read(old_dentry->d_inode));
5106 + if (err)
5107 + continue;
5108 + lower_new_dentry =
5109 + create_parents(dir, new_dentry,
5110 + new_dentry->d_name.name,
5111 + bindex);
5112 + lower_old_dentry = unionfs_lower_dentry(old_dentry);
5113 + lower_dir_dentry = lock_parent(lower_new_dentry);
5114 + /* see Documentation/filesystems/unionfs/issues.txt */
5115 + lockdep_off();
5116 + /* do vfs_link */
5117 + err = vfs_link(lower_old_dentry,
5118 + lower_dir_dentry->d_inode,
5119 + lower_new_dentry);
5120 + lockdep_on();
5121 + unlock_dir(lower_dir_dentry);
5122 + goto check_link;
5123 + }
5124 + goto out;
5125 + }
5126 +
5127 +check_link:
5128 + if (err || !lower_new_dentry->d_inode)
5129 + goto out;
5130 +
5131 + /* Its a hard link, so use the same inode */
5132 + new_dentry->d_inode = igrab(old_dentry->d_inode);
5133 + d_add(new_dentry, new_dentry->d_inode);
5134 + unionfs_copy_attr_all(dir, lower_new_dentry->d_parent->d_inode);
5135 + fsstack_copy_inode_size(dir, lower_new_dentry->d_parent->d_inode);
5136 +
5137 + /* propagate number of hard-links */
5138 + old_dentry->d_inode->i_nlink = unionfs_get_nlinks(old_dentry->d_inode);
5139 + /* new dentry's ctime may have changed due to hard-link counts */
5140 + unionfs_copy_attr_times(new_dentry->d_inode);
5141 +
5142 +out:
5143 + if (!new_dentry->d_inode)
5144 + d_drop(new_dentry);
5145 +
5146 + kfree(name);
5147 + if (!err)
5148 + unionfs_postcopyup_setmnt(new_dentry);
5149 +
5150 + unionfs_check_inode(dir);
5151 + unionfs_check_dentry(new_dentry);
5152 + unionfs_check_dentry(old_dentry);
5153 +
5154 + unionfs_double_unlock_dentry(old_dentry, new_dentry);
5155 + unionfs_double_unlock_parents(old_parent, new_parent);
5156 + dput(new_parent);
5157 + dput(old_parent);
5158 + unionfs_read_unlock(old_dentry->d_sb);
5159 +
5160 + return err;
5161 +}
5162 +
5163 +static int unionfs_symlink(struct inode *dir, struct dentry *dentry,
5164 + const char *symname)
5165 +{
5166 + int err = 0;
5167 + struct dentry *lower_dentry = NULL;
5168 + struct dentry *wh_dentry = NULL;
5169 + struct dentry *lower_parent_dentry = NULL;
5170 + struct dentry *parent;
5171 + char *name = NULL;
5172 + int valid = 0;
5173 + umode_t mode;
5174 +
5175 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5176 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
5177 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5178 +
5179 + valid = __unionfs_d_revalidate(dentry, parent, false);
5180 + if (unlikely(!valid)) {
5181 + err = -ESTALE;
5182 + goto out;
5183 + }
5184 +
5185 + /*
5186 + * It's only a bug if this dentry was not negative and couldn't be
5187 + * revalidated (shouldn't happen).
5188 + */
5189 + BUG_ON(!valid && dentry->d_inode);
5190 +
5191 + lower_dentry = find_writeable_branch(dir, dentry);
5192 + if (IS_ERR(lower_dentry)) {
5193 + err = PTR_ERR(lower_dentry);
5194 + goto out;
5195 + }
5196 +
5197 + lower_parent_dentry = lock_parent(lower_dentry);
5198 + if (IS_ERR(lower_parent_dentry)) {
5199 + err = PTR_ERR(lower_parent_dentry);
5200 + goto out;
5201 + }
5202 +
5203 + mode = S_IALLUGO;
5204 + err = vfs_symlink(lower_parent_dentry->d_inode, lower_dentry, symname);
5205 + if (!err) {
5206 + err = PTR_ERR(unionfs_interpose(dentry, dir->i_sb, 0));
5207 + if (!err) {
5208 + unionfs_copy_attr_times(dir);
5209 + fsstack_copy_inode_size(dir,
5210 + lower_parent_dentry->d_inode);
5211 + /* update no. of links on parent directory */
5212 + dir->i_nlink = unionfs_get_nlinks(dir);
5213 + }
5214 + }
5215 +
5216 + unlock_dir(lower_parent_dentry);
5217 +
5218 +out:
5219 + dput(wh_dentry);
5220 + kfree(name);
5221 +
5222 + if (!err) {
5223 + unionfs_postcopyup_setmnt(dentry);
5224 + unionfs_check_inode(dir);
5225 + unionfs_check_dentry(dentry);
5226 + }
5227 + unionfs_unlock_dentry(dentry);
5228 + unionfs_unlock_parent(dentry, parent);
5229 + unionfs_read_unlock(dentry->d_sb);
5230 + return err;
5231 +}
5232 +
5233 +static int unionfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
5234 +{
5235 + int err = 0;
5236 + struct dentry *lower_dentry = NULL;
5237 + struct dentry *lower_parent_dentry = NULL;
5238 + struct dentry *parent;
5239 + int bindex = 0, bstart;
5240 + char *name = NULL;
5241 + int valid;
5242 +
5243 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5244 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
5245 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5246 +
5247 + valid = __unionfs_d_revalidate(dentry, parent, false);
5248 + if (unlikely(!valid)) {
5249 + err = -ESTALE; /* same as what real_lookup does */
5250 + goto out;
5251 + }
5252 +
5253 + bstart = dbstart(dentry);
5254 +
5255 + lower_dentry = unionfs_lower_dentry(dentry);
5256 +
5257 + /* check for a whiteout in new dentry branch, and delete it */
5258 + err = check_unlink_whiteout(dentry, lower_dentry, bstart);
5259 + if (err > 0) /* whiteout found and removed successfully */
5260 + err = 0;
5261 + if (err) {
5262 + /* exit if the error returned was NOT -EROFS */
5263 + if (!IS_COPYUP_ERR(err))
5264 + goto out;
5265 + bstart--;
5266 + }
5267 +
5268 + /* check if copyup's needed, and mkdir */
5269 + for (bindex = bstart; bindex >= 0; bindex--) {
5270 + int i;
5271 + int bend = dbend(dentry);
5272 +
5273 + if (is_robranch_super(dentry->d_sb, bindex))
5274 + continue;
5275 +
5276 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
5277 + if (!lower_dentry) {
5278 + lower_dentry = create_parents(dir, dentry,
5279 + dentry->d_name.name,
5280 + bindex);
5281 + if (!lower_dentry || IS_ERR(lower_dentry)) {
5282 + printk(KERN_ERR "unionfs: lower dentry "
5283 + " NULL for bindex = %d\n", bindex);
5284 + continue;
5285 + }
5286 + }
5287 +
5288 + lower_parent_dentry = lock_parent(lower_dentry);
5289 +
5290 + if (IS_ERR(lower_parent_dentry)) {
5291 + err = PTR_ERR(lower_parent_dentry);
5292 + goto out;
5293 + }
5294 +
5295 + err = vfs_mkdir(lower_parent_dentry->d_inode, lower_dentry,
5296 + mode);
5297 +
5298 + unlock_dir(lower_parent_dentry);
5299 +
5300 + /* did the mkdir succeed? */
5301 + if (err)
5302 + break;
5303 +
5304 + for (i = bindex + 1; i <= bend; i++) {
5305 + /* XXX: use path_put_lowers? */
5306 + if (unionfs_lower_dentry_idx(dentry, i)) {
5307 + dput(unionfs_lower_dentry_idx(dentry, i));
5308 + unionfs_set_lower_dentry_idx(dentry, i, NULL);
5309 + }
5310 + }
5311 + dbend(dentry) = bindex;
5312 +
5313 + /*
5314 + * Only INTERPOSE_LOOKUP can return a value other than 0 on
5315 + * err.
5316 + */
5317 + err = PTR_ERR(unionfs_interpose(dentry, dir->i_sb, 0));
5318 + if (!err) {
5319 + unionfs_copy_attr_times(dir);
5320 + fsstack_copy_inode_size(dir,
5321 + lower_parent_dentry->d_inode);
5322 +
5323 + /* update number of links on parent directory */
5324 + dir->i_nlink = unionfs_get_nlinks(dir);
5325 + }
5326 +
5327 + err = make_dir_opaque(dentry, dbstart(dentry));
5328 + if (err) {
5329 + printk(KERN_ERR "unionfs: mkdir: error creating "
5330 + ".wh.__dir_opaque: %d\n", err);
5331 + goto out;
5332 + }
5333 +
5334 + /* we are done! */
5335 + break;
5336 + }
5337 +
5338 +out:
5339 + if (!dentry->d_inode)
5340 + d_drop(dentry);
5341 +
5342 + kfree(name);
5343 +
5344 + if (!err) {
5345 + unionfs_copy_attr_times(dentry->d_inode);
5346 + unionfs_postcopyup_setmnt(dentry);
5347 + }
5348 + unionfs_check_inode(dir);
5349 + unionfs_check_dentry(dentry);
5350 + unionfs_unlock_dentry(dentry);
5351 + unionfs_unlock_parent(dentry, parent);
5352 + unionfs_read_unlock(dentry->d_sb);
5353 +
5354 + return err;
5355 +}
5356 +
5357 +static int unionfs_mknod(struct inode *dir, struct dentry *dentry, int mode,
5358 + dev_t dev)
5359 +{
5360 + int err = 0;
5361 + struct dentry *lower_dentry = NULL;
5362 + struct dentry *wh_dentry = NULL;
5363 + struct dentry *lower_parent_dentry = NULL;
5364 + struct dentry *parent;
5365 + char *name = NULL;
5366 + int valid = 0;
5367 +
5368 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5369 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
5370 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5371 +
5372 + valid = __unionfs_d_revalidate(dentry, parent, false);
5373 + if (unlikely(!valid)) {
5374 + err = -ESTALE;
5375 + goto out;
5376 + }
5377 +
5378 + /*
5379 + * It's only a bug if this dentry was not negative and couldn't be
5380 + * revalidated (shouldn't happen).
5381 + */
5382 + BUG_ON(!valid && dentry->d_inode);
5383 +
5384 + lower_dentry = find_writeable_branch(dir, dentry);
5385 + if (IS_ERR(lower_dentry)) {
5386 + err = PTR_ERR(lower_dentry);
5387 + goto out;
5388 + }
5389 +
5390 + lower_parent_dentry = lock_parent(lower_dentry);
5391 + if (IS_ERR(lower_parent_dentry)) {
5392 + err = PTR_ERR(lower_parent_dentry);
5393 + goto out;
5394 + }
5395 +
5396 + err = vfs_mknod(lower_parent_dentry->d_inode, lower_dentry, mode, dev);
5397 + if (!err) {
5398 + err = PTR_ERR(unionfs_interpose(dentry, dir->i_sb, 0));
5399 + if (!err) {
5400 + unionfs_copy_attr_times(dir);
5401 + fsstack_copy_inode_size(dir,
5402 + lower_parent_dentry->d_inode);
5403 + /* update no. of links on parent directory */
5404 + dir->i_nlink = unionfs_get_nlinks(dir);
5405 + }
5406 + }
5407 +
5408 + unlock_dir(lower_parent_dentry);
5409 +
5410 +out:
5411 + dput(wh_dentry);
5412 + kfree(name);
5413 +
5414 + if (!err) {
5415 + unionfs_postcopyup_setmnt(dentry);
5416 + unionfs_check_inode(dir);
5417 + unionfs_check_dentry(dentry);
5418 + }
5419 + unionfs_unlock_dentry(dentry);
5420 + unionfs_unlock_parent(dentry, parent);
5421 + unionfs_read_unlock(dentry->d_sb);
5422 + return err;
5423 +}
5424 +
5425 +/* requires sb, dentry, and parent to already be locked */
5426 +static int __unionfs_readlink(struct dentry *dentry, char __user *buf,
5427 + int bufsiz)
5428 +{
5429 + int err;
5430 + struct dentry *lower_dentry;
5431 +
5432 + lower_dentry = unionfs_lower_dentry(dentry);
5433 +
5434 + if (!lower_dentry->d_inode->i_op ||
5435 + !lower_dentry->d_inode->i_op->readlink) {
5436 + err = -EINVAL;
5437 + goto out;
5438 + }
5439 +
5440 + err = lower_dentry->d_inode->i_op->readlink(lower_dentry,
5441 + buf, bufsiz);
5442 + if (err >= 0)
5443 + fsstack_copy_attr_atime(dentry->d_inode,
5444 + lower_dentry->d_inode);
5445 +
5446 +out:
5447 + return err;
5448 +}
5449 +
5450 +static int unionfs_readlink(struct dentry *dentry, char __user *buf,
5451 + int bufsiz)
5452 +{
5453 + int err;
5454 + struct dentry *parent;
5455 +
5456 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5457 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
5458 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5459 +
5460 + if (unlikely(!__unionfs_d_revalidate(dentry, parent, false))) {
5461 + err = -ESTALE;
5462 + goto out;
5463 + }
5464 +
5465 + err = __unionfs_readlink(dentry, buf, bufsiz);
5466 +
5467 +out:
5468 + unionfs_check_dentry(dentry);
5469 + unionfs_unlock_dentry(dentry);
5470 + unionfs_unlock_parent(dentry, parent);
5471 + unionfs_read_unlock(dentry->d_sb);
5472 +
5473 + return err;
5474 +}
5475 +
5476 +static void *unionfs_follow_link(struct dentry *dentry, struct nameidata *nd)
5477 +{
5478 + char *buf;
5479 + int len = PAGE_SIZE, err;
5480 + mm_segment_t old_fs;
5481 + struct dentry *parent;
5482 +
5483 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5484 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
5485 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5486 +
5487 + /* This is freed by the put_link method assuming a successful call. */
5488 + buf = kmalloc(len, GFP_KERNEL);
5489 + if (unlikely(!buf)) {
5490 + err = -ENOMEM;
5491 + goto out;
5492 + }
5493 +
5494 + /* read the symlink, and then we will follow it */
5495 + old_fs = get_fs();
5496 + set_fs(KERNEL_DS);
5497 + err = __unionfs_readlink(dentry, buf, len);
5498 + set_fs(old_fs);
5499 + if (err < 0) {
5500 + kfree(buf);
5501 + buf = NULL;
5502 + goto out;
5503 + }
5504 + buf[err] = 0;
5505 + nd_set_link(nd, buf);
5506 + err = 0;
5507 +
5508 +out:
5509 + if (err >= 0) {
5510 + unionfs_check_nd(nd);
5511 + unionfs_check_dentry(dentry);
5512 + }
5513 +
5514 + unionfs_unlock_dentry(dentry);
5515 + unionfs_unlock_parent(dentry, parent);
5516 + unionfs_read_unlock(dentry->d_sb);
5517 +
5518 + return ERR_PTR(err);
5519 +}
5520 +
5521 +/* this @nd *IS* still used */
5522 +static void unionfs_put_link(struct dentry *dentry, struct nameidata *nd,
5523 + void *cookie)
5524 +{
5525 + struct dentry *parent;
5526 +
5527 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5528 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
5529 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5530 +
5531 + if (unlikely(!__unionfs_d_revalidate(dentry, parent, false)))
5532 + printk(KERN_ERR
5533 + "unionfs: put_link failed to revalidate dentry\n");
5534 +
5535 + unionfs_check_dentry(dentry);
5536 + unionfs_check_nd(nd);
5537 + kfree(nd_get_link(nd));
5538 + unionfs_unlock_dentry(dentry);
5539 + unionfs_unlock_parent(dentry, parent);
5540 + unionfs_read_unlock(dentry->d_sb);
5541 +}
5542 +
5543 +/*
5544 + * This is a variant of fs/namei.c:permission() or inode_permission() which
5545 + * skips over EROFS tests (because we perform copyup on EROFS).
5546 + */
5547 +static int __inode_permission(struct inode *inode, int mask)
5548 +{
5549 + int retval;
5550 +
5551 + /* nobody gets write access to an immutable file */
5552 + if ((mask & MAY_WRITE) && IS_IMMUTABLE(inode))
5553 + return -EACCES;
5554 +
5555 + /* Ordinary permission routines do not understand MAY_APPEND. */
5556 + if (inode->i_op && inode->i_op->permission) {
5557 + retval = inode->i_op->permission(inode, mask);
5558 + if (!retval) {
5559 + /*
5560 + * Exec permission on a regular file is denied if none
5561 + * of the execute bits are set.
5562 + *
5563 + * This check should be done by the ->permission()
5564 + * method.
5565 + */
5566 + if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode) &&
5567 + !(inode->i_mode & S_IXUGO))
5568 + return -EACCES;
5569 + }
5570 + } else {
5571 + retval = generic_permission(inode, mask, NULL);
5572 + }
5573 + if (retval)
5574 + return retval;
5575 +
5576 + return security_inode_permission(inode,
5577 + mask & (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND));
5578 +}
5579 +
5580 +/*
5581 + * Don't grab the superblock read-lock in unionfs_permission, which prevents
5582 + * a deadlock with the branch-management "add branch" code (which grabbed
5583 + * the write lock). It is safe to not grab the read lock here, because even
5584 + * with branch management taking place, there is no chance that
5585 + * unionfs_permission, or anything it calls, will use stale branch
5586 + * information.
5587 + */
5588 +static int unionfs_permission(struct inode *inode, int mask)
5589 +{
5590 + struct inode *lower_inode = NULL;
5591 + int err = 0;
5592 + int bindex, bstart, bend;
5593 + const int is_file = !S_ISDIR(inode->i_mode);
5594 + const int write_mask = (mask & MAY_WRITE) && !(mask & MAY_READ);
5595 + struct inode *inode_grabbed = igrab(inode);
5596 + struct dentry *dentry = d_find_alias(inode);
5597 +
5598 + if (dentry)
5599 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5600 +
5601 + if (!UNIONFS_I(inode)->lower_inodes) {
5602 + if (is_file) /* dirs can be unlinked but chdir'ed to */
5603 + err = -ESTALE; /* force revalidate */
5604 + goto out;
5605 + }
5606 + bstart = ibstart(inode);
5607 + bend = ibend(inode);
5608 + if (unlikely(bstart < 0 || bend < 0)) {
5609 + /*
5610 + * With branch-management, we can get a stale inode here.
5611 + * If so, we return ESTALE back to link_path_walk, which
5612 + * would discard the dcache entry and re-lookup the
5613 + * dentry+inode. This should be equivalent to issuing
5614 + * __unionfs_d_revalidate_chain on nd.dentry here.
5615 + */
5616 + if (is_file) /* dirs can be unlinked but chdir'ed to */
5617 + err = -ESTALE; /* force revalidate */
5618 + goto out;
5619 + }
5620 +
5621 + for (bindex = bstart; bindex <= bend; bindex++) {
5622 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
5623 + if (!lower_inode)
5624 + continue;
5625 +
5626 + /*
5627 + * check the condition for D-F-D underlying files/directories,
5628 + * we don't have to check for files, if we are checking for
5629 + * directories.
5630 + */
5631 + if (!is_file && !S_ISDIR(lower_inode->i_mode))
5632 + continue;
5633 +
5634 + /*
5635 + * We check basic permissions, but we ignore any conditions
5636 + * such as readonly file systems or branches marked as
5637 + * readonly, because those conditions should lead to a
5638 + * copyup taking place later on. However, if user never had
5639 + * access to the file, then no copyup could ever take place.
5640 + */
5641 + err = __inode_permission(lower_inode, mask);
5642 + if (err && err != -EACCES && err != EPERM && bindex > 0) {
5643 + umode_t mode = lower_inode->i_mode;
5644 + if ((is_robranch_super(inode->i_sb, bindex) ||
5645 + IS_RDONLY(lower_inode)) &&
5646 + (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
5647 + err = 0;
5648 + if (IS_COPYUP_ERR(err))
5649 + err = 0;
5650 + }
5651 +
5652 + /*
5653 + * The permissions are an intersection of the overall directory
5654 + * permissions, so we fail if one fails.
5655 + */
5656 + if (err)
5657 + goto out;
5658 +
5659 + /* only the leftmost file matters. */
5660 + if (is_file || write_mask) {
5661 + if (is_file && write_mask) {
5662 + err = get_write_access(lower_inode);
5663 + if (!err)
5664 + put_write_access(lower_inode);
5665 + }
5666 + break;
5667 + }
5668 + }
5669 + /* sync times which may have changed (asynchronously) below */
5670 + unionfs_copy_attr_times(inode);
5671 +
5672 +out:
5673 + unionfs_check_inode(inode);
5674 + if (dentry) {
5675 + unionfs_unlock_dentry(dentry);
5676 + dput(dentry);
5677 + }
5678 + iput(inode_grabbed);
5679 + return err;
5680 +}
5681 +
5682 +static int unionfs_setattr(struct dentry *dentry, struct iattr *ia)
5683 +{
5684 + int err = 0;
5685 + struct dentry *lower_dentry;
5686 + struct dentry *parent;
5687 + struct inode *inode;
5688 + struct inode *lower_inode;
5689 + int bstart, bend, bindex;
5690 + loff_t size;
5691 +
5692 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
5693 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
5694 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
5695 +
5696 + if (unlikely(!__unionfs_d_revalidate(dentry, parent, false))) {
5697 + err = -ESTALE;
5698 + goto out;
5699 + }
5700 +
5701 + bstart = dbstart(dentry);
5702 + bend = dbend(dentry);
5703 + inode = dentry->d_inode;
5704 +
5705 + /*
5706 + * mode change is for clearing setuid/setgid. Allow lower filesystem
5707 + * to reinterpret it in its own way.
5708 + */
5709 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
5710 + ia->ia_valid &= ~ATTR_MODE;
5711 +
5712 + lower_dentry = unionfs_lower_dentry(dentry);
5713 + if (!lower_dentry) { /* should never happen after above revalidate */
5714 + err = -EINVAL;
5715 + goto out;
5716 + }
5717 + lower_inode = unionfs_lower_inode(inode);
5718 +
5719 + /* check if user has permission to change lower inode */
5720 + err = inode_change_ok(lower_inode, ia);
5721 + if (err)
5722 + goto out;
5723 +
5724 + /* copyup if the file is on a read only branch */
5725 + if (is_robranch_super(dentry->d_sb, bstart)
5726 + || IS_RDONLY(lower_inode)) {
5727 + /* check if we have a branch to copy up to */
5728 + if (bstart <= 0) {
5729 + err = -EACCES;
5730 + goto out;
5731 + }
5732 +
5733 + if (ia->ia_valid & ATTR_SIZE)
5734 + size = ia->ia_size;
5735 + else
5736 + size = i_size_read(inode);
5737 + /* copyup to next available branch */
5738 + for (bindex = bstart - 1; bindex >= 0; bindex--) {
5739 + err = copyup_dentry(parent->d_inode,
5740 + dentry, bstart, bindex,
5741 + dentry->d_name.name,
5742 + dentry->d_name.len,
5743 + NULL, size);
5744 + if (!err)
5745 + break;
5746 + }
5747 + if (err)
5748 + goto out;
5749 + /* get updated lower_dentry/inode after copyup */
5750 + lower_dentry = unionfs_lower_dentry(dentry);
5751 + lower_inode = unionfs_lower_inode(inode);
5752 + }
5753 +
5754 + /*
5755 + * If shrinking, first truncate upper level to cancel writing dirty
5756 + * pages beyond the new eof; and also if its' maxbytes is more
5757 + * limiting (fail with -EFBIG before making any change to the lower
5758 + * level). There is no need to vmtruncate the upper level
5759 + * afterwards in the other cases: we fsstack_copy_inode_size from
5760 + * the lower level.
5761 + */
5762 + if (ia->ia_valid & ATTR_SIZE) {
5763 + size = i_size_read(inode);
5764 + if (ia->ia_size < size || (ia->ia_size > size &&
5765 + inode->i_sb->s_maxbytes < lower_inode->i_sb->s_maxbytes)) {
5766 + err = vmtruncate(inode, ia->ia_size);
5767 + if (err)
5768 + goto out;
5769 + }
5770 + }
5771 +
5772 + /* notify the (possibly copied-up) lower inode */
5773 + mutex_lock(&lower_inode->i_mutex);
5774 + err = notify_change(lower_dentry, ia);
5775 + mutex_unlock(&lower_inode->i_mutex);
5776 + if (err)
5777 + goto out;
5778 +
5779 + /* get attributes from the first lower inode */
5780 + unionfs_copy_attr_all(inode, lower_inode);
5781 + /*
5782 + * unionfs_copy_attr_all will copy the lower times to our inode if
5783 + * the lower ones are newer (useful for cache coherency). However,
5784 + * ->setattr is the only place in which we may have to copy the
5785 + * lower inode times absolutely, to support utimes(2).
5786 + */
5787 + if (ia->ia_valid & ATTR_MTIME_SET)
5788 + inode->i_mtime = lower_inode->i_mtime;
5789 + if (ia->ia_valid & ATTR_CTIME)
5790 + inode->i_ctime = lower_inode->i_ctime;
5791 + if (ia->ia_valid & ATTR_ATIME_SET)
5792 + inode->i_atime = lower_inode->i_atime;
5793 + fsstack_copy_inode_size(inode, lower_inode);
5794 +
5795 +out:
5796 + if (!err)
5797 + unionfs_check_dentry(dentry);
5798 + unionfs_unlock_dentry(dentry);
5799 + unionfs_unlock_parent(dentry, parent);
5800 + unionfs_read_unlock(dentry->d_sb);
5801 +
5802 + return err;
5803 +}
5804 +
5805 +struct inode_operations unionfs_symlink_iops = {
5806 + .readlink = unionfs_readlink,
5807 + .permission = unionfs_permission,
5808 + .follow_link = unionfs_follow_link,
5809 + .setattr = unionfs_setattr,
5810 + .put_link = unionfs_put_link,
5811 +};
5812 +
5813 +struct inode_operations unionfs_dir_iops = {
5814 + .create = unionfs_create,
5815 + .lookup = unionfs_lookup,
5816 + .link = unionfs_link,
5817 + .unlink = unionfs_unlink,
5818 + .symlink = unionfs_symlink,
5819 + .mkdir = unionfs_mkdir,
5820 + .rmdir = unionfs_rmdir,
5821 + .mknod = unionfs_mknod,
5822 + .rename = unionfs_rename,
5823 + .permission = unionfs_permission,
5824 + .setattr = unionfs_setattr,
5825 +#ifdef CONFIG_UNION_FS_XATTR
5826 + .setxattr = unionfs_setxattr,
5827 + .getxattr = unionfs_getxattr,
5828 + .removexattr = unionfs_removexattr,
5829 + .listxattr = unionfs_listxattr,
5830 +#endif /* CONFIG_UNION_FS_XATTR */
5831 +};
5832 +
5833 +struct inode_operations unionfs_main_iops = {
5834 + .permission = unionfs_permission,
5835 + .setattr = unionfs_setattr,
5836 +#ifdef CONFIG_UNION_FS_XATTR
5837 + .setxattr = unionfs_setxattr,
5838 + .getxattr = unionfs_getxattr,
5839 + .removexattr = unionfs_removexattr,
5840 + .listxattr = unionfs_listxattr,
5841 +#endif /* CONFIG_UNION_FS_XATTR */
5842 +};
5843 diff -Naur linux-2.6.30/fs/unionfs/lookup.c linux-2.6.30-magellan/fs/unionfs/lookup.c
5844 --- linux-2.6.30/fs/unionfs/lookup.c 1970-01-01 01:00:00.000000000 +0100
5845 +++ linux-2.6.30-magellan/fs/unionfs/lookup.c 2009-06-11 18:39:19.000000000 +0200
5846 @@ -0,0 +1,569 @@
5847 +/*
5848 + * Copyright (c) 2003-2009 Erez Zadok
5849 + * Copyright (c) 2003-2006 Charles P. Wright
5850 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
5851 + * Copyright (c) 2005-2006 Junjiro Okajima
5852 + * Copyright (c) 2005 Arun M. Krishnakumar
5853 + * Copyright (c) 2004-2006 David P. Quigley
5854 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
5855 + * Copyright (c) 2003 Puja Gupta
5856 + * Copyright (c) 2003 Harikesavan Krishnan
5857 + * Copyright (c) 2003-2009 Stony Brook University
5858 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
5859 + *
5860 + * This program is free software; you can redistribute it and/or modify
5861 + * it under the terms of the GNU General Public License version 2 as
5862 + * published by the Free Software Foundation.
5863 + */
5864 +
5865 +#include "union.h"
5866 +
5867 +/*
5868 + * Lookup one path component @name relative to a <base,mnt> path pair.
5869 + * Behaves nearly the same as lookup_one_len (i.e., return negative dentry
5870 + * on ENOENT), but uses the @mnt passed, so it can cross bind mounts and
5871 + * other lower mounts properly. If @new_mnt is non-null, will fill in the
5872 + * new mnt there. Caller is responsible to dput/mntput/path_put returned
5873 + * @dentry and @new_mnt.
5874 + */
5875 +struct dentry *__lookup_one(struct dentry *base, struct vfsmount *mnt,
5876 + const char *name, struct vfsmount **new_mnt)
5877 +{
5878 + struct dentry *dentry = NULL;
5879 + struct nameidata lower_nd;
5880 + int err;
5881 +
5882 + /* we use flags=0 to get basic lookup */
5883 + err = vfs_path_lookup(base, mnt, name, 0, &lower_nd);
5884 +
5885 + switch (err) {
5886 + case 0: /* no error */
5887 + dentry = lower_nd.path.dentry;
5888 + if (new_mnt)
5889 + *new_mnt = lower_nd.path.mnt; /* rc already inc'ed */
5890 + break;
5891 + case -ENOENT:
5892 + /*
5893 + * We don't consider ENOENT an error, and we want to return
5894 + * a negative dentry (ala lookup_one_len). As we know
5895 + * there was no inode for this name before (-ENOENT), then
5896 + * it's safe to call lookup_one_len (which doesn't take a
5897 + * vfsmount).
5898 + */
5899 + dentry = lookup_one_len(name, base, strlen(name));
5900 + if (new_mnt)
5901 + *new_mnt = mntget(lower_nd.path.mnt);
5902 + break;
5903 + default: /* all other real errors */
5904 + dentry = ERR_PTR(err);
5905 + break;
5906 + }
5907 +
5908 + return dentry;
5909 +}
5910 +
5911 +/*
5912 + * This is a utility function that fills in a unionfs dentry.
5913 + * Caller must lock this dentry with unionfs_lock_dentry.
5914 + *
5915 + * Returns: 0 (ok), or -ERRNO if an error occurred.
5916 + * XXX: get rid of _partial_lookup and make callers call _lookup_full directly
5917 + */
5918 +int unionfs_partial_lookup(struct dentry *dentry, struct dentry *parent)
5919 +{
5920 + struct dentry *tmp;
5921 + int err = -ENOSYS;
5922 +
5923 + tmp = unionfs_lookup_full(dentry, parent, INTERPOSE_PARTIAL);
5924 +
5925 + if (!tmp) {
5926 + err = 0;
5927 + goto out;
5928 + }
5929 + if (IS_ERR(tmp)) {
5930 + err = PTR_ERR(tmp);
5931 + goto out;
5932 + }
5933 + /* XXX: need to change the interface */
5934 + BUG_ON(tmp != dentry);
5935 +out:
5936 + return err;
5937 +}
5938 +
5939 +/* The dentry cache is just so we have properly sized dentries. */
5940 +static struct kmem_cache *unionfs_dentry_cachep;
5941 +int unionfs_init_dentry_cache(void)
5942 +{
5943 + unionfs_dentry_cachep =
5944 + kmem_cache_create("unionfs_dentry",
5945 + sizeof(struct unionfs_dentry_info),
5946 + 0, SLAB_RECLAIM_ACCOUNT, NULL);
5947 +
5948 + return (unionfs_dentry_cachep ? 0 : -ENOMEM);
5949 +}
5950 +
5951 +void unionfs_destroy_dentry_cache(void)
5952 +{
5953 + if (unionfs_dentry_cachep)
5954 + kmem_cache_destroy(unionfs_dentry_cachep);
5955 +}
5956 +
5957 +void free_dentry_private_data(struct dentry *dentry)
5958 +{
5959 + if (!dentry || !dentry->d_fsdata)
5960 + return;
5961 + kfree(UNIONFS_D(dentry)->lower_paths);
5962 + UNIONFS_D(dentry)->lower_paths = NULL;
5963 + kmem_cache_free(unionfs_dentry_cachep, dentry->d_fsdata);
5964 + dentry->d_fsdata = NULL;
5965 +}
5966 +
5967 +static inline int __realloc_dentry_private_data(struct dentry *dentry)
5968 +{
5969 + struct unionfs_dentry_info *info = UNIONFS_D(dentry);
5970 + void *p;
5971 + int size;
5972 +
5973 + BUG_ON(!info);
5974 +
5975 + size = sizeof(struct path) * sbmax(dentry->d_sb);
5976 + p = krealloc(info->lower_paths, size, GFP_ATOMIC);
5977 + if (unlikely(!p))
5978 + return -ENOMEM;
5979 +
5980 + info->lower_paths = p;
5981 +
5982 + info->bstart = -1;
5983 + info->bend = -1;
5984 + info->bopaque = -1;
5985 + info->bcount = sbmax(dentry->d_sb);
5986 + atomic_set(&info->generation,
5987 + atomic_read(&UNIONFS_SB(dentry->d_sb)->generation));
5988 +
5989 + memset(info->lower_paths, 0, size);
5990 +
5991 + return 0;
5992 +}
5993 +
5994 +/* UNIONFS_D(dentry)->lock must be locked */
5995 +int realloc_dentry_private_data(struct dentry *dentry)
5996 +{
5997 + if (!__realloc_dentry_private_data(dentry))
5998 + return 0;
5999 +
6000 + kfree(UNIONFS_D(dentry)->lower_paths);
6001 + free_dentry_private_data(dentry);
6002 + return -ENOMEM;
6003 +}
6004 +
6005 +/* allocate new dentry private data */
6006 +int new_dentry_private_data(struct dentry *dentry, int subclass)
6007 +{
6008 + struct unionfs_dentry_info *info = UNIONFS_D(dentry);
6009 +
6010 + BUG_ON(info);
6011 +
6012 + info = kmem_cache_alloc(unionfs_dentry_cachep, GFP_ATOMIC);
6013 + if (unlikely(!info))
6014 + return -ENOMEM;
6015 +
6016 + mutex_init(&info->lock);
6017 + mutex_lock_nested(&info->lock, subclass);
6018 +
6019 + info->lower_paths = NULL;
6020 +
6021 + dentry->d_fsdata = info;
6022 +
6023 + if (!__realloc_dentry_private_data(dentry))
6024 + return 0;
6025 +
6026 + mutex_unlock(&info->lock);
6027 + free_dentry_private_data(dentry);
6028 + return -ENOMEM;
6029 +}
6030 +
6031 +/*
6032 + * scan through the lower dentry objects, and set bstart to reflect the
6033 + * starting branch
6034 + */
6035 +void update_bstart(struct dentry *dentry)
6036 +{
6037 + int bindex;
6038 + int bstart = dbstart(dentry);
6039 + int bend = dbend(dentry);
6040 + struct dentry *lower_dentry;
6041 +
6042 + for (bindex = bstart; bindex <= bend; bindex++) {
6043 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
6044 + if (!lower_dentry)
6045 + continue;
6046 + if (lower_dentry->d_inode) {
6047 + dbstart(dentry) = bindex;
6048 + break;
6049 + }
6050 + dput(lower_dentry);
6051 + unionfs_set_lower_dentry_idx(dentry, bindex, NULL);
6052 + }
6053 +}
6054 +
6055 +
6056 +/*
6057 + * Initialize a nameidata structure (the intent part) we can pass to a lower
6058 + * file system. Returns 0 on success or -error (only -ENOMEM possible).
6059 + * Inside that nd structure, this function may also return an allocated
6060 + * struct file (for open intents). The caller, when done with this nd, must
6061 + * kfree the intent file (using release_lower_nd).
6062 + *
6063 + * XXX: this code, and the callers of this code, should be redone using
6064 + * vfs_path_lookup() when (1) the nameidata structure is refactored into a
6065 + * separate intent-structure, and (2) open_namei() is broken into a VFS-only
6066 + * function and a method that other file systems can call.
6067 + */
6068 +int init_lower_nd(struct nameidata *nd, unsigned int flags)
6069 +{
6070 + int err = 0;
6071 +#ifdef ALLOC_LOWER_ND_FILE
6072 + /*
6073 + * XXX: one day we may need to have the lower return an open file
6074 + * for us. It is not needed in 2.6.23-rc1 for nfs2/nfs3, but may
6075 + * very well be needed for nfs4.
6076 + */
6077 + struct file *file;
6078 +#endif /* ALLOC_LOWER_ND_FILE */
6079 +
6080 + memset(nd, 0, sizeof(struct nameidata));
6081 + if (!flags)
6082 + return err;
6083 +
6084 + switch (flags) {
6085 + case LOOKUP_CREATE:
6086 + nd->intent.open.flags |= O_CREAT;
6087 + /* fall through: shared code for create/open cases */
6088 + case LOOKUP_OPEN:
6089 + nd->flags = flags;
6090 + nd->intent.open.flags |= (FMODE_READ | FMODE_WRITE);
6091 +#ifdef ALLOC_LOWER_ND_FILE
6092 + file = kzalloc(sizeof(struct file), GFP_KERNEL);
6093 + if (unlikely(!file)) {
6094 + err = -ENOMEM;
6095 + break; /* exit switch statement and thus return */
6096 + }
6097 + nd->intent.open.file = file;
6098 +#endif /* ALLOC_LOWER_ND_FILE */
6099 + break;
6100 + default:
6101 + /*
6102 + * We should never get here, for now.
6103 + * We can add new cases here later on.
6104 + */
6105 + pr_debug("unionfs: unknown nameidata flag 0x%x\n", flags);
6106 + BUG();
6107 + break;
6108 + }
6109 +
6110 + return err;
6111 +}
6112 +
6113 +void release_lower_nd(struct nameidata *nd, int err)
6114 +{
6115 + if (!nd->intent.open.file)
6116 + return;
6117 + else if (!err)
6118 + release_open_intent(nd);
6119 +#ifdef ALLOC_LOWER_ND_FILE
6120 + kfree(nd->intent.open.file);
6121 +#endif /* ALLOC_LOWER_ND_FILE */
6122 +}
6123 +
6124 +/*
6125 + * Main (and complex) driver function for Unionfs's lookup
6126 + *
6127 + * Returns: NULL (ok), ERR_PTR if an error occurred, or a non-null non-error
6128 + * PTR if d_splice returned a different dentry.
6129 + *
6130 + * If lookupmode is INTERPOSE_PARTIAL/REVAL/REVAL_NEG, the passed dentry's
6131 + * inode info must be locked. If lookupmode is INTERPOSE_LOOKUP (i.e., a
6132 + * newly looked-up dentry), then unionfs_lookup_backend will return a locked
6133 + * dentry's info, which the caller must unlock.
6134 + */
6135 +struct dentry *unionfs_lookup_full(struct dentry *dentry,
6136 + struct dentry *parent, int lookupmode)
6137 +{
6138 + int err = 0;
6139 + struct dentry *lower_dentry = NULL;
6140 + struct vfsmount *lower_mnt;
6141 + struct vfsmount *lower_dir_mnt;
6142 + struct dentry *wh_lower_dentry = NULL;
6143 + struct dentry *lower_dir_dentry = NULL;
6144 + struct dentry *d_interposed = NULL;
6145 + int bindex, bstart, bend, bopaque;
6146 + int opaque, num_positive = 0;
6147 + const char *name;
6148 + int namelen;
6149 + int pos_start, pos_end;
6150 +
6151 + /*
6152 + * We should already have a lock on this dentry in the case of a
6153 + * partial lookup, or a revalidation. Otherwise it is returned from
6154 + * new_dentry_private_data already locked.
6155 + */
6156 + verify_locked(dentry);
6157 + verify_locked(parent);
6158 +
6159 + /* must initialize dentry operations */
6160 + dentry->d_op = &unionfs_dops;
6161 +
6162 + /* We never partial lookup the root directory. */
6163 + if (IS_ROOT(dentry))
6164 + goto out;
6165 +
6166 + name = dentry->d_name.name;
6167 + namelen = dentry->d_name.len;
6168 +
6169 + /* No dentries should get created for possible whiteout names. */
6170 + if (!is_validname(name)) {
6171 + err = -EPERM;
6172 + goto out_free;
6173 + }
6174 +
6175 + /* Now start the actual lookup procedure. */
6176 + bstart = dbstart(parent);
6177 + bend = dbend(parent);
6178 + bopaque = dbopaque(parent);
6179 + BUG_ON(bstart < 0);
6180 +
6181 + /* adjust bend to bopaque if needed */
6182 + if ((bopaque >= 0) && (bopaque < bend))
6183 + bend = bopaque;
6184 +
6185 + /* lookup all possible dentries */
6186 + for (bindex = bstart; bindex <= bend; bindex++) {
6187 +
6188 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
6189 + lower_mnt = unionfs_lower_mnt_idx(dentry, bindex);
6190 +
6191 + /* skip if we already have a positive lower dentry */
6192 + if (lower_dentry) {
6193 + if (dbstart(dentry) < 0)
6194 + dbstart(dentry) = bindex;
6195 + if (bindex > dbend(dentry))
6196 + dbend(dentry) = bindex;
6197 + if (lower_dentry->d_inode)
6198 + num_positive++;
6199 + continue;
6200 + }
6201 +
6202 + lower_dir_dentry =
6203 + unionfs_lower_dentry_idx(parent, bindex);
6204 + /* if the lower dentry's parent does not exist, skip this */
6205 + if (!lower_dir_dentry || !lower_dir_dentry->d_inode)
6206 + continue;
6207 +
6208 + /* also skip it if the parent isn't a directory. */
6209 + if (!S_ISDIR(lower_dir_dentry->d_inode->i_mode))
6210 + continue; /* XXX: should be BUG_ON */
6211 +
6212 + /* check for whiteouts: stop lookup if found */
6213 + wh_lower_dentry = lookup_whiteout(name, lower_dir_dentry);
6214 + if (IS_ERR(wh_lower_dentry)) {
6215 + err = PTR_ERR(wh_lower_dentry);
6216 + goto out_free;
6217 + }
6218 + if (wh_lower_dentry->d_inode) {
6219 + dbend(dentry) = dbopaque(dentry) = bindex;
6220 + if (dbstart(dentry) < 0)
6221 + dbstart(dentry) = bindex;
6222 + dput(wh_lower_dentry);
6223 + break;
6224 + }
6225 + dput(wh_lower_dentry);
6226 +
6227 + /* Now do regular lookup; lookup @name */
6228 + lower_dir_mnt = unionfs_lower_mnt_idx(parent, bindex);
6229 + lower_mnt = NULL; /* XXX: needed? */
6230 +
6231 + lower_dentry = __lookup_one(lower_dir_dentry, lower_dir_mnt,
6232 + name, &lower_mnt);
6233 +
6234 + if (IS_ERR(lower_dentry)) {
6235 + err = PTR_ERR(lower_dentry);
6236 + goto out_free;
6237 + }
6238 + unionfs_set_lower_dentry_idx(dentry, bindex, lower_dentry);
6239 + if (!lower_mnt)
6240 + lower_mnt = unionfs_mntget(dentry->d_sb->s_root,
6241 + bindex);
6242 + unionfs_set_lower_mnt_idx(dentry, bindex, lower_mnt);
6243 +
6244 + /* adjust dbstart/end */
6245 + if (dbstart(dentry) < 0)
6246 + dbstart(dentry) = bindex;
6247 + if (bindex > dbend(dentry))
6248 + dbend(dentry) = bindex;
6249 + /*
6250 + * We always store the lower dentries above, and update
6251 + * dbstart/dbend, even if the whole unionfs dentry is
6252 + * negative (i.e., no lower inodes).
6253 + */
6254 + if (!lower_dentry->d_inode)
6255 + continue;
6256 + num_positive++;
6257 +
6258 + /*
6259 + * check if we just found an opaque directory, if so, stop
6260 + * lookups here.
6261 + */
6262 + if (!S_ISDIR(lower_dentry->d_inode->i_mode))
6263 + continue;
6264 + opaque = is_opaque_dir(dentry, bindex);
6265 + if (opaque < 0) {
6266 + err = opaque;
6267 + goto out_free;
6268 + } else if (opaque) {
6269 + dbend(dentry) = dbopaque(dentry) = bindex;
6270 + break;
6271 + }
6272 + dbend(dentry) = bindex;
6273 +
6274 + /* update parent directory's atime with the bindex */
6275 + fsstack_copy_attr_atime(parent->d_inode,
6276 + lower_dir_dentry->d_inode);
6277 + }
6278 +
6279 + /* sanity checks, then decide if to process a negative dentry */
6280 + BUG_ON(dbstart(dentry) < 0 && dbend(dentry) >= 0);
6281 + BUG_ON(dbstart(dentry) >= 0 && dbend(dentry) < 0);
6282 +
6283 + if (num_positive > 0)
6284 + goto out_positive;
6285 +
6286 + /*** handle NEGATIVE dentries ***/
6287 +
6288 + /*
6289 + * If negative, keep only first lower negative dentry, to save on
6290 + * memory.
6291 + */
6292 + if (dbstart(dentry) < dbend(dentry)) {
6293 + path_put_lowers(dentry, dbstart(dentry) + 1,
6294 + dbend(dentry), false);
6295 + dbend(dentry) = dbstart(dentry);
6296 + }
6297 + if (lookupmode == INTERPOSE_PARTIAL)
6298 + goto out;
6299 + if (lookupmode == INTERPOSE_LOOKUP) {
6300 + /*
6301 + * If all we found was a whiteout in the first available
6302 + * branch, then create a negative dentry for a possibly new
6303 + * file to be created.
6304 + */
6305 + if (dbopaque(dentry) < 0)
6306 + goto out;
6307 + /* XXX: need to get mnt here */
6308 + bindex = dbstart(dentry);
6309 + if (unionfs_lower_dentry_idx(dentry, bindex))
6310 + goto out;
6311 + lower_dir_dentry =
6312 + unionfs_lower_dentry_idx(parent, bindex);
6313 + if (!lower_dir_dentry || !lower_dir_dentry->d_inode)
6314 + goto out;
6315 + if (!S_ISDIR(lower_dir_dentry->d_inode->i_mode))
6316 + goto out; /* XXX: should be BUG_ON */
6317 + /* XXX: do we need to cross bind mounts here? */
6318 + lower_dentry = lookup_one_len(name, lower_dir_dentry, namelen);
6319 + if (IS_ERR(lower_dentry)) {
6320 + err = PTR_ERR(lower_dentry);
6321 + goto out;
6322 + }
6323 + /* XXX: need to mntget/mntput as needed too! */
6324 + unionfs_set_lower_dentry_idx(dentry, bindex, lower_dentry);
6325 + /* XXX: wrong mnt for crossing bind mounts! */
6326 + lower_mnt = unionfs_mntget(dentry->d_sb->s_root, bindex);
6327 + unionfs_set_lower_mnt_idx(dentry, bindex, lower_mnt);
6328 +
6329 + goto out;
6330 + }
6331 +
6332 + /* if we're revalidating a positive dentry, don't make it negative */
6333 + if (lookupmode != INTERPOSE_REVAL)
6334 + d_add(dentry, NULL);
6335 +
6336 + goto out;
6337 +
6338 +out_positive:
6339 + /*** handle POSITIVE dentries ***/
6340 +
6341 + /*
6342 + * This unionfs dentry is positive (at least one lower inode
6343 + * exists), so scan entire dentry from beginning to end, and remove
6344 + * any negative lower dentries, if any. Then, update dbstart/dbend
6345 + * to reflect the start/end of positive dentries.
6346 + */
6347 + pos_start = pos_end = -1;
6348 + for (bindex = bstart; bindex <= bend; bindex++) {
6349 + lower_dentry = unionfs_lower_dentry_idx(dentry,
6350 + bindex);
6351 + if (lower_dentry && lower_dentry->d_inode) {
6352 + if (pos_start < 0)
6353 + pos_start = bindex;
6354 + if (bindex > pos_end)
6355 + pos_end = bindex;
6356 + continue;
6357 + }
6358 + path_put_lowers(dentry, bindex, bindex, false);
6359 + }
6360 + if (pos_start >= 0)
6361 + dbstart(dentry) = pos_start;
6362 + if (pos_end >= 0)
6363 + dbend(dentry) = pos_end;
6364 +
6365 + /* Partial lookups need to re-interpose, or throw away older negs. */
6366 + if (lookupmode == INTERPOSE_PARTIAL) {
6367 + if (dentry->d_inode) {
6368 + unionfs_reinterpose(dentry);
6369 + goto out;
6370 + }
6371 +
6372 + /*
6373 + * This dentry was positive, so it is as if we had a
6374 + * negative revalidation.
6375 + */
6376 + lookupmode = INTERPOSE_REVAL_NEG;
6377 + update_bstart(dentry);
6378 + }
6379 +
6380 + /*
6381 + * Interpose can return a dentry if d_splice returned a different
6382 + * dentry.
6383 + */
6384 + d_interposed = unionfs_interpose(dentry, dentry->d_sb, lookupmode);
6385 + if (IS_ERR(d_interposed))
6386 + err = PTR_ERR(d_interposed);
6387 + else if (d_interposed)
6388 + dentry = d_interposed;
6389 +
6390 + if (!err)
6391 + goto out;
6392 + d_drop(dentry);
6393 +
6394 +out_free:
6395 + /* should dput/mntput all the underlying dentries on error condition */
6396 + if (dbstart(dentry) >= 0)
6397 + path_put_lowers_all(dentry, false);
6398 + /* free lower_paths unconditionally */
6399 + kfree(UNIONFS_D(dentry)->lower_paths);
6400 + UNIONFS_D(dentry)->lower_paths = NULL;
6401 +
6402 +out:
6403 + if (dentry && UNIONFS_D(dentry)) {
6404 + BUG_ON(dbstart(dentry) < 0 && dbend(dentry) >= 0);
6405 + BUG_ON(dbstart(dentry) >= 0 && dbend(dentry) < 0);
6406 + }
6407 + if (d_interposed && UNIONFS_D(d_interposed)) {
6408 + BUG_ON(dbstart(d_interposed) < 0 && dbend(d_interposed) >= 0);
6409 + BUG_ON(dbstart(d_interposed) >= 0 && dbend(d_interposed) < 0);
6410 + }
6411 +
6412 + if (!err && d_interposed)
6413 + return d_interposed;
6414 + return ERR_PTR(err);
6415 +}
6416 diff -Naur linux-2.6.30/fs/unionfs/main.c linux-2.6.30-magellan/fs/unionfs/main.c
6417 --- linux-2.6.30/fs/unionfs/main.c 1970-01-01 01:00:00.000000000 +0100
6418 +++ linux-2.6.30-magellan/fs/unionfs/main.c 2009-06-11 18:39:19.000000000 +0200
6419 @@ -0,0 +1,758 @@
6420 +/*
6421 + * Copyright (c) 2003-2009 Erez Zadok
6422 + * Copyright (c) 2003-2006 Charles P. Wright
6423 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
6424 + * Copyright (c) 2005-2006 Junjiro Okajima
6425 + * Copyright (c) 2005 Arun M. Krishnakumar
6426 + * Copyright (c) 2004-2006 David P. Quigley
6427 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
6428 + * Copyright (c) 2003 Puja Gupta
6429 + * Copyright (c) 2003 Harikesavan Krishnan
6430 + * Copyright (c) 2003-2009 Stony Brook University
6431 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
6432 + *
6433 + * This program is free software; you can redistribute it and/or modify
6434 + * it under the terms of the GNU General Public License version 2 as
6435 + * published by the Free Software Foundation.
6436 + */
6437 +
6438 +#include "union.h"
6439 +#include <linux/module.h>
6440 +#include <linux/moduleparam.h>
6441 +
6442 +static void unionfs_fill_inode(struct dentry *dentry,
6443 + struct inode *inode)
6444 +{
6445 + struct inode *lower_inode;
6446 + struct dentry *lower_dentry;
6447 + int bindex, bstart, bend;
6448 +
6449 + bstart = dbstart(dentry);
6450 + bend = dbend(dentry);
6451 +
6452 + for (bindex = bstart; bindex <= bend; bindex++) {
6453 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
6454 + if (!lower_dentry) {
6455 + unionfs_set_lower_inode_idx(inode, bindex, NULL);
6456 + continue;
6457 + }
6458 +
6459 + /* Initialize the lower inode to the new lower inode. */
6460 + if (!lower_dentry->d_inode)
6461 + continue;
6462 +
6463 + unionfs_set_lower_inode_idx(inode, bindex,
6464 + igrab(lower_dentry->d_inode));
6465 + }
6466 +
6467 + ibstart(inode) = dbstart(dentry);
6468 + ibend(inode) = dbend(dentry);
6469 +
6470 + /* Use attributes from the first branch. */
6471 + lower_inode = unionfs_lower_inode(inode);
6472 +
6473 + /* Use different set of inode ops for symlinks & directories */
6474 + if (S_ISLNK(lower_inode->i_mode))
6475 + inode->i_op = &unionfs_symlink_iops;
6476 + else if (S_ISDIR(lower_inode->i_mode))
6477 + inode->i_op = &unionfs_dir_iops;
6478 +
6479 + /* Use different set of file ops for directories */
6480 + if (S_ISDIR(lower_inode->i_mode))
6481 + inode->i_fop = &unionfs_dir_fops;
6482 +
6483 + /* properly initialize special inodes */
6484 + if (S_ISBLK(lower_inode->i_mode) || S_ISCHR(lower_inode->i_mode) ||
6485 + S_ISFIFO(lower_inode->i_mode) || S_ISSOCK(lower_inode->i_mode))
6486 + init_special_inode(inode, lower_inode->i_mode,
6487 + lower_inode->i_rdev);
6488 +
6489 + /* all well, copy inode attributes */
6490 + unionfs_copy_attr_all(inode, lower_inode);
6491 + fsstack_copy_inode_size(inode, lower_inode);
6492 +}
6493 +
6494 +/*
6495 + * Connect a unionfs inode dentry/inode with several lower ones. This is
6496 + * the classic stackable file system "vnode interposition" action.
6497 + *
6498 + * @sb: unionfs's super_block
6499 + */
6500 +struct dentry *unionfs_interpose(struct dentry *dentry, struct super_block *sb,
6501 + int flag)
6502 +{
6503 + int err = 0;
6504 + struct inode *inode;
6505 + int need_fill_inode = 1;
6506 + struct dentry *spliced = NULL;
6507 +
6508 + verify_locked(dentry);
6509 +
6510 + /*
6511 + * We allocate our new inode below by calling unionfs_iget,
6512 + * which will initialize some of the new inode's fields
6513 + */
6514 +
6515 + /*
6516 + * On revalidate we've already got our own inode and just need
6517 + * to fix it up.
6518 + */
6519 + if (flag == INTERPOSE_REVAL) {
6520 + inode = dentry->d_inode;
6521 + UNIONFS_I(inode)->bstart = -1;
6522 + UNIONFS_I(inode)->bend = -1;
6523 + atomic_set(&UNIONFS_I(inode)->generation,
6524 + atomic_read(&UNIONFS_SB(sb)->generation));
6525 +
6526 + UNIONFS_I(inode)->lower_inodes =
6527 + kcalloc(sbmax(sb), sizeof(struct inode *), GFP_KERNEL);
6528 + if (unlikely(!UNIONFS_I(inode)->lower_inodes)) {
6529 + err = -ENOMEM;
6530 + goto out;
6531 + }
6532 + } else {
6533 + /* get unique inode number for unionfs */
6534 + inode = unionfs_iget(sb, iunique(sb, UNIONFS_ROOT_INO));
6535 + if (IS_ERR(inode)) {
6536 + err = PTR_ERR(inode);
6537 + goto out;
6538 + }
6539 + if (atomic_read(&inode->i_count) > 1)
6540 + goto skip;
6541 + }
6542 +
6543 + need_fill_inode = 0;
6544 + unionfs_fill_inode(dentry, inode);
6545 +
6546 +skip:
6547 + /* only (our) lookup wants to do a d_add */
6548 + switch (flag) {
6549 + case INTERPOSE_DEFAULT:
6550 + /* for operations which create new inodes */
6551 + d_add(dentry, inode);
6552 + break;
6553 + case INTERPOSE_REVAL_NEG:
6554 + d_instantiate(dentry, inode);
6555 + break;
6556 + case INTERPOSE_LOOKUP:
6557 + spliced = d_splice_alias(inode, dentry);
6558 + if (spliced && spliced != dentry) {
6559 + /*
6560 + * d_splice can return a dentry if it was
6561 + * disconnected and had to be moved. We must ensure
6562 + * that the private data of the new dentry is
6563 + * correct and that the inode info was filled
6564 + * properly. Finally we must return this new
6565 + * dentry.
6566 + */
6567 + spliced->d_op = &unionfs_dops;
6568 + spliced->d_fsdata = dentry->d_fsdata;
6569 + dentry->d_fsdata = NULL;
6570 + dentry = spliced;
6571 + if (need_fill_inode) {
6572 + need_fill_inode = 0;
6573 + unionfs_fill_inode(dentry, inode);
6574 + }
6575 + goto out_spliced;
6576 + } else if (!spliced) {
6577 + if (need_fill_inode) {
6578 + need_fill_inode = 0;
6579 + unionfs_fill_inode(dentry, inode);
6580 + goto out_spliced;
6581 + }
6582 + }
6583 + break;
6584 + case INTERPOSE_REVAL:
6585 + /* Do nothing. */
6586 + break;
6587 + default:
6588 + printk(KERN_CRIT "unionfs: invalid interpose flag passed!\n");
6589 + BUG();
6590 + }
6591 + goto out;
6592 +
6593 +out_spliced:
6594 + if (!err)
6595 + return spliced;
6596 +out:
6597 + return ERR_PTR(err);
6598 +}
6599 +
6600 +/* like interpose above, but for an already existing dentry */
6601 +void unionfs_reinterpose(struct dentry *dentry)
6602 +{
6603 + struct dentry *lower_dentry;
6604 + struct inode *inode;
6605 + int bindex, bstart, bend;
6606 +
6607 + verify_locked(dentry);
6608 +
6609 + /* This is pre-allocated inode */
6610 + inode = dentry->d_inode;
6611 +
6612 + bstart = dbstart(dentry);
6613 + bend = dbend(dentry);
6614 + for (bindex = bstart; bindex <= bend; bindex++) {
6615 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
6616 + if (!lower_dentry)
6617 + continue;
6618 +
6619 + if (!lower_dentry->d_inode)
6620 + continue;
6621 + if (unionfs_lower_inode_idx(inode, bindex))
6622 + continue;
6623 + unionfs_set_lower_inode_idx(inode, bindex,
6624 + igrab(lower_dentry->d_inode));
6625 + }
6626 + ibstart(inode) = dbstart(dentry);
6627 + ibend(inode) = dbend(dentry);
6628 +}
6629 +
6630 +/*
6631 + * make sure the branch we just looked up (nd) makes sense:
6632 + *
6633 + * 1) we're not trying to stack unionfs on top of unionfs
6634 + * 2) it exists
6635 + * 3) is a directory
6636 + */
6637 +int check_branch(struct nameidata *nd)
6638 +{
6639 + /* XXX: remove in ODF code -- stacking unions allowed there */
6640 + if (!strcmp(nd->path.dentry->d_sb->s_type->name, UNIONFS_NAME))
6641 + return -EINVAL;
6642 + if (!nd->path.dentry->d_inode)
6643 + return -ENOENT;
6644 + if (!S_ISDIR(nd->path.dentry->d_inode->i_mode))
6645 + return -ENOTDIR;
6646 + return 0;
6647 +}
6648 +
6649 +/* checks if two lower_dentries have overlapping branches */
6650 +static int is_branch_overlap(struct dentry *dent1, struct dentry *dent2)
6651 +{
6652 + struct dentry *dent = NULL;
6653 +
6654 + dent = dent1;
6655 + while ((dent != dent2) && (dent->d_parent != dent))
6656 + dent = dent->d_parent;
6657 +
6658 + if (dent == dent2)
6659 + return 1;
6660 +
6661 + dent = dent2;
6662 + while ((dent != dent1) && (dent->d_parent != dent))
6663 + dent = dent->d_parent;
6664 +
6665 + return (dent == dent1);
6666 +}
6667 +
6668 +/*
6669 + * Parse "ro" or "rw" options, but default to "rw" if no mode options was
6670 + * specified. Fill the mode bits in @perms. If encounter an unknown
6671 + * string, return -EINVAL. Otherwise return 0.
6672 + */
6673 +int parse_branch_mode(const char *name, int *perms)
6674 +{
6675 + if (!name || !strcmp(name, "rw")) {
6676 + *perms = MAY_READ | MAY_WRITE;
6677 + return 0;
6678 + }
6679 + if (!strcmp(name, "ro")) {
6680 + *perms = MAY_READ;
6681 + return 0;
6682 + }
6683 + return -EINVAL;
6684 +}
6685 +
6686 +/*
6687 + * parse the dirs= mount argument
6688 + *
6689 + * We don't need to lock the superblock private data's rwsem, as we get
6690 + * called only by unionfs_read_super - it is still a long time before anyone
6691 + * can even get a reference to us.
6692 + */
6693 +static int parse_dirs_option(struct super_block *sb, struct unionfs_dentry_info
6694 + *lower_root_info, char *options)
6695 +{
6696 + struct nameidata nd;
6697 + char *name;
6698 + int err = 0;
6699 + int branches = 1;
6700 + int bindex = 0;
6701 + int i = 0;
6702 + int j = 0;
6703 + struct dentry *dent1;
6704 + struct dentry *dent2;
6705 +
6706 + if (options[0] == '\0') {
6707 + printk(KERN_ERR "unionfs: no branches specified\n");
6708 + err = -EINVAL;
6709 + goto out;
6710 + }
6711 +
6712 + /*
6713 + * Each colon means we have a separator, this is really just a rough
6714 + * guess, since strsep will handle empty fields for us.
6715 + */
6716 + for (i = 0; options[i]; i++)
6717 + if (options[i] == ':')
6718 + branches++;
6719 +
6720 + /* allocate space for underlying pointers to lower dentry */
6721 + UNIONFS_SB(sb)->data =
6722 + kcalloc(branches, sizeof(struct unionfs_data), GFP_KERNEL);
6723 + if (unlikely(!UNIONFS_SB(sb)->data)) {
6724 + err = -ENOMEM;
6725 + goto out;
6726 + }
6727 +
6728 + lower_root_info->lower_paths =
6729 + kcalloc(branches, sizeof(struct path), GFP_KERNEL);
6730 + if (unlikely(!lower_root_info->lower_paths)) {
6731 + err = -ENOMEM;
6732 + goto out;
6733 + }
6734 +
6735 + /* now parsing a string such as "b1:b2=rw:b3=ro:b4" */
6736 + branches = 0;
6737 + while ((name = strsep(&options, ":")) != NULL) {
6738 + int perms;
6739 + char *mode = strchr(name, '=');
6740 +
6741 + if (!name)
6742 + continue;
6743 + if (!*name) { /* bad use of ':' (extra colons) */
6744 + err = -EINVAL;
6745 + goto out;
6746 + }
6747 +
6748 + branches++;
6749 +
6750 + /* strip off '=' if any */
6751 + if (mode)
6752 + *mode++ = '\0';
6753 +
6754 + err = parse_branch_mode(mode, &perms);
6755 + if (err) {
6756 + printk(KERN_ERR "unionfs: invalid mode \"%s\" for "
6757 + "branch %d\n", mode, bindex);
6758 + goto out;
6759 + }
6760 + /* ensure that leftmost branch is writeable */
6761 + if (!bindex && !(perms & MAY_WRITE)) {
6762 + printk(KERN_ERR "unionfs: leftmost branch cannot be "
6763 + "read-only (use \"-o ro\" to create a "
6764 + "read-only union)\n");
6765 + err = -EINVAL;
6766 + goto out;
6767 + }
6768 +
6769 + err = path_lookup(name, LOOKUP_FOLLOW, &nd);
6770 + if (err) {
6771 + printk(KERN_ERR "unionfs: error accessing "
6772 + "lower directory '%s' (error %d)\n",
6773 + name, err);
6774 + goto out;
6775 + }
6776 +
6777 + err = check_branch(&nd);
6778 + if (err) {
6779 + printk(KERN_ERR "unionfs: lower directory "
6780 + "'%s' is not a valid branch\n", name);
6781 + path_put(&nd.path);
6782 + goto out;
6783 + }
6784 +
6785 + lower_root_info->lower_paths[bindex].dentry = nd.path.dentry;
6786 + lower_root_info->lower_paths[bindex].mnt = nd.path.mnt;
6787 +
6788 + set_branchperms(sb, bindex, perms);
6789 + set_branch_count(sb, bindex, 0);
6790 + new_branch_id(sb, bindex);
6791 +
6792 + if (lower_root_info->bstart < 0)
6793 + lower_root_info->bstart = bindex;
6794 + lower_root_info->bend = bindex;
6795 + bindex++;
6796 + }
6797 +
6798 + if (branches == 0) {
6799 + printk(KERN_ERR "unionfs: no branches specified\n");
6800 + err = -EINVAL;
6801 + goto out;
6802 + }
6803 +
6804 + BUG_ON(branches != (lower_root_info->bend + 1));
6805 +
6806 + /*
6807 + * Ensure that no overlaps exist in the branches.
6808 + *
6809 + * This test is required because the Linux kernel has no support
6810 + * currently for ensuring coherency between stackable layers and
6811 + * branches. If we were to allow overlapping branches, it would be
6812 + * possible, for example, to delete a file via one branch, which
6813 + * would not be reflected in another branch. Such incoherency could
6814 + * lead to inconsistencies and even kernel oopses. Rather than
6815 + * implement hacks to work around some of these cache-coherency
6816 + * problems, we prevent branch overlapping, for now. A complete
6817 + * solution will involve proper kernel/VFS support for cache
6818 + * coherency, at which time we could safely remove this
6819 + * branch-overlapping test.
6820 + */
6821 + for (i = 0; i < branches; i++) {
6822 + dent1 = lower_root_info->lower_paths[i].dentry;
6823 + for (j = i + 1; j < branches; j++) {
6824 + dent2 = lower_root_info->lower_paths[j].dentry;
6825 + if (is_branch_overlap(dent1, dent2)) {
6826 + printk(KERN_ERR "unionfs: branches %d and "
6827 + "%d overlap\n", i, j);
6828 + err = -EINVAL;
6829 + goto out;
6830 + }
6831 + }
6832 + }
6833 +
6834 +out:
6835 + if (err) {
6836 + for (i = 0; i < branches; i++)
6837 + path_put(&lower_root_info->lower_paths[i]);
6838 +
6839 + kfree(lower_root_info->lower_paths);
6840 + kfree(UNIONFS_SB(sb)->data);
6841 +
6842 + /*
6843 + * MUST clear the pointers to prevent potential double free if
6844 + * the caller dies later on
6845 + */
6846 + lower_root_info->lower_paths = NULL;
6847 + UNIONFS_SB(sb)->data = NULL;
6848 + }
6849 + return err;
6850 +}
6851 +
6852 +/*
6853 + * Parse mount options. See the manual page for usage instructions.
6854 + *
6855 + * Returns the dentry object of the lower-level (lower) directory;
6856 + * We want to mount our stackable file system on top of that lower directory.
6857 + */
6858 +static struct unionfs_dentry_info *unionfs_parse_options(
6859 + struct super_block *sb,
6860 + char *options)
6861 +{
6862 + struct unionfs_dentry_info *lower_root_info;
6863 + char *optname;
6864 + int err = 0;
6865 + int bindex;
6866 + int dirsfound = 0;
6867 +
6868 + /* allocate private data area */
6869 + err = -ENOMEM;
6870 + lower_root_info =
6871 + kzalloc(sizeof(struct unionfs_dentry_info), GFP_KERNEL);
6872 + if (unlikely(!lower_root_info))
6873 + goto out_error;
6874 + lower_root_info->bstart = -1;
6875 + lower_root_info->bend = -1;
6876 + lower_root_info->bopaque = -1;
6877 +
6878 + while ((optname = strsep(&options, ",")) != NULL) {
6879 + char *optarg;
6880 +
6881 + if (!optname || !*optname)
6882 + continue;
6883 +
6884 + optarg = strchr(optname, '=');
6885 + if (optarg)
6886 + *optarg++ = '\0';
6887 +
6888 + /*
6889 + * All of our options take an argument now. Insert ones that
6890 + * don't, above this check.
6891 + */
6892 + if (!optarg) {
6893 + printk(KERN_ERR "unionfs: %s requires an argument\n",
6894 + optname);
6895 + err = -EINVAL;
6896 + goto out_error;
6897 + }
6898 +
6899 + if (!strcmp("dirs", optname)) {
6900 + if (++dirsfound > 1) {
6901 + printk(KERN_ERR
6902 + "unionfs: multiple dirs specified\n");
6903 + err = -EINVAL;
6904 + goto out_error;
6905 + }
6906 + err = parse_dirs_option(sb, lower_root_info, optarg);
6907 + if (err)
6908 + goto out_error;
6909 + continue;
6910 + }
6911 +
6912 + err = -EINVAL;
6913 + printk(KERN_ERR
6914 + "unionfs: unrecognized option '%s'\n", optname);
6915 + goto out_error;
6916 + }
6917 + if (dirsfound != 1) {
6918 + printk(KERN_ERR "unionfs: dirs option required\n");
6919 + err = -EINVAL;
6920 + goto out_error;
6921 + }
6922 + goto out;
6923 +
6924 +out_error:
6925 + if (lower_root_info && lower_root_info->lower_paths) {
6926 + for (bindex = lower_root_info->bstart;
6927 + bindex >= 0 && bindex <= lower_root_info->bend;
6928 + bindex++)
6929 + path_put(&lower_root_info->lower_paths[bindex]);
6930 + }
6931 +
6932 + kfree(lower_root_info->lower_paths);
6933 + kfree(lower_root_info);
6934 +
6935 + kfree(UNIONFS_SB(sb)->data);
6936 + UNIONFS_SB(sb)->data = NULL;
6937 +
6938 + lower_root_info = ERR_PTR(err);
6939 +out:
6940 + return lower_root_info;
6941 +}
6942 +
6943 +/*
6944 + * our custom d_alloc_root work-alike
6945 + *
6946 + * we can't use d_alloc_root if we want to use our own interpose function
6947 + * unchanged, so we simply call our own "fake" d_alloc_root
6948 + */
6949 +static struct dentry *unionfs_d_alloc_root(struct super_block *sb)
6950 +{
6951 + struct dentry *ret = NULL;
6952 +
6953 + if (sb) {
6954 + static const struct qstr name = {
6955 + .name = "/",
6956 + .len = 1
6957 + };
6958 +
6959 + ret = d_alloc(NULL, &name);
6960 + if (likely(ret)) {
6961 + ret->d_op = &unionfs_dops;
6962 + ret->d_sb = sb;
6963 + ret->d_parent = ret;
6964 + }
6965 + }
6966 + return ret;
6967 +}
6968 +
6969 +/*
6970 + * There is no need to lock the unionfs_super_info's rwsem as there is no
6971 + * way anyone can have a reference to the superblock at this point in time.
6972 + */
6973 +static int unionfs_read_super(struct super_block *sb, void *raw_data,
6974 + int silent)
6975 +{
6976 + int err = 0;
6977 + struct unionfs_dentry_info *lower_root_info = NULL;
6978 + int bindex, bstart, bend;
6979 +
6980 + if (!raw_data) {
6981 + printk(KERN_ERR
6982 + "unionfs: read_super: missing data argument\n");
6983 + err = -EINVAL;
6984 + goto out;
6985 + }
6986 +
6987 + /* Allocate superblock private data */
6988 + sb->s_fs_info = kzalloc(sizeof(struct unionfs_sb_info), GFP_KERNEL);
6989 + if (unlikely(!UNIONFS_SB(sb))) {
6990 + printk(KERN_CRIT "unionfs: read_super: out of memory\n");
6991 + err = -ENOMEM;
6992 + goto out;
6993 + }
6994 +
6995 + UNIONFS_SB(sb)->bend = -1;
6996 + atomic_set(&UNIONFS_SB(sb)->generation, 1);
6997 + init_rwsem(&UNIONFS_SB(sb)->rwsem);
6998 + UNIONFS_SB(sb)->high_branch_id = -1; /* -1 == invalid branch ID */
6999 +
7000 + lower_root_info = unionfs_parse_options(sb, raw_data);
7001 + if (IS_ERR(lower_root_info)) {
7002 + printk(KERN_ERR
7003 + "unionfs: read_super: error while parsing options "
7004 + "(err = %ld)\n", PTR_ERR(lower_root_info));
7005 + err = PTR_ERR(lower_root_info);
7006 + lower_root_info = NULL;
7007 + goto out_free;
7008 + }
7009 + if (lower_root_info->bstart == -1) {
7010 + err = -ENOENT;
7011 + goto out_free;
7012 + }
7013 +
7014 + /* set the lower superblock field of upper superblock */
7015 + bstart = lower_root_info->bstart;
7016 + BUG_ON(bstart != 0);
7017 + sbend(sb) = bend = lower_root_info->bend;
7018 + for (bindex = bstart; bindex <= bend; bindex++) {
7019 + struct dentry *d = lower_root_info->lower_paths[bindex].dentry;
7020 + atomic_inc(&d->d_sb->s_active);
7021 + unionfs_set_lower_super_idx(sb, bindex, d->d_sb);
7022 + }
7023 +
7024 + /* max Bytes is the maximum bytes from highest priority branch */
7025 + sb->s_maxbytes = unionfs_lower_super_idx(sb, 0)->s_maxbytes;
7026 +
7027 + /*
7028 + * Our c/m/atime granularity is 1 ns because we may stack on file
7029 + * systems whose granularity is as good. This is important for our
7030 + * time-based cache coherency.
7031 + */
7032 + sb->s_time_gran = 1;
7033 +
7034 + sb->s_op = &unionfs_sops;
7035 +
7036 + /* See comment next to the definition of unionfs_d_alloc_root */
7037 + sb->s_root = unionfs_d_alloc_root(sb);
7038 + if (unlikely(!sb->s_root)) {
7039 + err = -ENOMEM;
7040 + goto out_dput;
7041 + }
7042 +
7043 + /* link the upper and lower dentries */
7044 + sb->s_root->d_fsdata = NULL;
7045 + err = new_dentry_private_data(sb->s_root, UNIONFS_DMUTEX_ROOT);
7046 + if (unlikely(err))
7047 + goto out_freedpd;
7048 +
7049 + /* Set the lower dentries for s_root */
7050 + for (bindex = bstart; bindex <= bend; bindex++) {
7051 + struct dentry *d;
7052 + struct vfsmount *m;
7053 +
7054 + d = lower_root_info->lower_paths[bindex].dentry;
7055 + m = lower_root_info->lower_paths[bindex].mnt;
7056 +
7057 + unionfs_set_lower_dentry_idx(sb->s_root, bindex, d);
7058 + unionfs_set_lower_mnt_idx(sb->s_root, bindex, m);
7059 + }
7060 + dbstart(sb->s_root) = bstart;
7061 + dbend(sb->s_root) = bend;
7062 +
7063 + /* Set the generation number to one, since this is for the mount. */
7064 + atomic_set(&UNIONFS_D(sb->s_root)->generation, 1);
7065 +
7066 + /*
7067 + * Call interpose to create the upper level inode. Only
7068 + * INTERPOSE_LOOKUP can return a value other than 0 on err.
7069 + */
7070 + err = PTR_ERR(unionfs_interpose(sb->s_root, sb, 0));
7071 + unionfs_unlock_dentry(sb->s_root);
7072 + if (!err)
7073 + goto out;
7074 + /* else fall through */
7075 +
7076 +out_freedpd:
7077 + if (UNIONFS_D(sb->s_root)) {
7078 + kfree(UNIONFS_D(sb->s_root)->lower_paths);
7079 + free_dentry_private_data(sb->s_root);
7080 + }
7081 + dput(sb->s_root);
7082 +
7083 +out_dput:
7084 + if (lower_root_info && !IS_ERR(lower_root_info)) {
7085 + for (bindex = lower_root_info->bstart;
7086 + bindex <= lower_root_info->bend; bindex++) {
7087 + struct dentry *d;
7088 + d = lower_root_info->lower_paths[bindex].dentry;
7089 + /* drop refs we took earlier */
7090 + atomic_dec(&d->d_sb->s_active);
7091 + path_put(&lower_root_info->lower_paths[bindex]);
7092 + }
7093 + kfree(lower_root_info->lower_paths);
7094 + kfree(lower_root_info);
7095 + lower_root_info = NULL;
7096 + }
7097 +
7098 +out_free:
7099 + kfree(UNIONFS_SB(sb)->data);
7100 + kfree(UNIONFS_SB(sb));
7101 + sb->s_fs_info = NULL;
7102 +
7103 +out:
7104 + if (lower_root_info && !IS_ERR(lower_root_info)) {
7105 + kfree(lower_root_info->lower_paths);
7106 + kfree(lower_root_info);
7107 + }
7108 + return err;
7109 +}
7110 +
7111 +static int unionfs_get_sb(struct file_system_type *fs_type,
7112 + int flags, const char *dev_name,
7113 + void *raw_data, struct vfsmount *mnt)
7114 +{
7115 + int err;
7116 + err = get_sb_nodev(fs_type, flags, raw_data, unionfs_read_super, mnt);
7117 + if (!err)
7118 + UNIONFS_SB(mnt->mnt_sb)->dev_name =
7119 + kstrdup(dev_name, GFP_KERNEL);
7120 + return err;
7121 +}
7122 +
7123 +static struct file_system_type unionfs_fs_type = {
7124 + .owner = THIS_MODULE,
7125 + .name = UNIONFS_NAME,
7126 + .get_sb = unionfs_get_sb,
7127 + .kill_sb = generic_shutdown_super,
7128 + .fs_flags = FS_REVAL_DOT,
7129 +};
7130 +
7131 +static int __init init_unionfs_fs(void)
7132 +{
7133 + int err;
7134 +
7135 + pr_info("Registering unionfs " UNIONFS_VERSION "\n");
7136 +
7137 + err = unionfs_init_filldir_cache();
7138 + if (unlikely(err))
7139 + goto out;
7140 + err = unionfs_init_inode_cache();
7141 + if (unlikely(err))
7142 + goto out;
7143 + err = unionfs_init_dentry_cache();
7144 + if (unlikely(err))
7145 + goto out;
7146 + err = init_sioq();
7147 + if (unlikely(err))
7148 + goto out;
7149 + err = register_filesystem(&unionfs_fs_type);
7150 +out:
7151 + if (unlikely(err)) {
7152 + stop_sioq();
7153 + unionfs_destroy_filldir_cache();
7154 + unionfs_destroy_inode_cache();
7155 + unionfs_destroy_dentry_cache();
7156 + }
7157 + return err;
7158 +}
7159 +
7160 +static void __exit exit_unionfs_fs(void)
7161 +{
7162 + stop_sioq();
7163 + unionfs_destroy_filldir_cache();
7164 + unionfs_destroy_inode_cache();
7165 + unionfs_destroy_dentry_cache();
7166 + unregister_filesystem(&unionfs_fs_type);
7167 + pr_info("Completed unionfs module unload\n");
7168 +}
7169 +
7170 +MODULE_AUTHOR("Erez Zadok, Filesystems and Storage Lab, Stony Brook University"
7171 + " (http://www.fsl.cs.sunysb.edu)");
7172 +MODULE_DESCRIPTION("Unionfs " UNIONFS_VERSION
7173 + " (http://unionfs.filesystems.org)");
7174 +MODULE_LICENSE("GPL");
7175 +
7176 +module_init(init_unionfs_fs);
7177 +module_exit(exit_unionfs_fs);
7178 diff -Naur linux-2.6.30/fs/unionfs/Makefile linux-2.6.30-magellan/fs/unionfs/Makefile
7179 --- linux-2.6.30/fs/unionfs/Makefile 1970-01-01 01:00:00.000000000 +0100
7180 +++ linux-2.6.30-magellan/fs/unionfs/Makefile 2009-06-11 18:39:19.000000000 +0200
7181 @@ -0,0 +1,17 @@
7182 +UNIONFS_VERSION="2.5.1 (for 2.6.29-rc2)"
7183 +
7184 +EXTRA_CFLAGS += -DUNIONFS_VERSION=\"$(UNIONFS_VERSION)\"
7185 +
7186 +obj-$(CONFIG_UNION_FS) += unionfs.o
7187 +
7188 +unionfs-y := subr.o dentry.o file.o inode.o main.o super.o \
7189 + rdstate.o copyup.o dirhelper.o rename.o unlink.o \
7190 + lookup.o commonfops.o dirfops.o sioq.o mmap.o whiteout.o
7191 +
7192 +unionfs-$(CONFIG_UNION_FS_XATTR) += xattr.o
7193 +
7194 +unionfs-$(CONFIG_UNION_FS_DEBUG) += debug.o
7195 +
7196 +ifeq ($(CONFIG_UNION_FS_DEBUG),y)
7197 +EXTRA_CFLAGS += -DDEBUG
7198 +endif
7199 diff -Naur linux-2.6.30/fs/unionfs/mmap.c linux-2.6.30-magellan/fs/unionfs/mmap.c
7200 --- linux-2.6.30/fs/unionfs/mmap.c 1970-01-01 01:00:00.000000000 +0100
7201 +++ linux-2.6.30-magellan/fs/unionfs/mmap.c 2009-06-11 18:39:19.000000000 +0200
7202 @@ -0,0 +1,89 @@
7203 +/*
7204 + * Copyright (c) 2003-2009 Erez Zadok
7205 + * Copyright (c) 2003-2006 Charles P. Wright
7206 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
7207 + * Copyright (c) 2005-2006 Junjiro Okajima
7208 + * Copyright (c) 2006 Shaya Potter
7209 + * Copyright (c) 2005 Arun M. Krishnakumar
7210 + * Copyright (c) 2004-2006 David P. Quigley
7211 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
7212 + * Copyright (c) 2003 Puja Gupta
7213 + * Copyright (c) 2003 Harikesavan Krishnan
7214 + * Copyright (c) 2003-2009 Stony Brook University
7215 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
7216 + *
7217 + * This program is free software; you can redistribute it and/or modify
7218 + * it under the terms of the GNU General Public License version 2 as
7219 + * published by the Free Software Foundation.
7220 + */
7221 +
7222 +#include "union.h"
7223 +
7224 +
7225 +/*
7226 + * XXX: we need a dummy readpage handler because generic_file_mmap (which we
7227 + * use in unionfs_mmap) checks for the existence of
7228 + * mapping->a_ops->readpage, else it returns -ENOEXEC. The VFS will need to
7229 + * be fixed to allow a file system to define vm_ops->fault without any
7230 + * address_space_ops whatsoever.
7231 + *
7232 + * Otherwise, we don't want to use our readpage method at all.
7233 + */
7234 +static int unionfs_readpage(struct file *file, struct page *page)
7235 +{
7236 + BUG();
7237 + return -EINVAL;
7238 +}
7239 +
7240 +static int unionfs_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
7241 +{
7242 + int err;
7243 + struct file *file, *lower_file;
7244 + struct vm_operations_struct *lower_vm_ops;
7245 + struct vm_area_struct lower_vma;
7246 +
7247 + BUG_ON(!vma);
7248 + memcpy(&lower_vma, vma, sizeof(struct vm_area_struct));
7249 + file = lower_vma.vm_file;
7250 + lower_vm_ops = UNIONFS_F(file)->lower_vm_ops;
7251 + BUG_ON(!lower_vm_ops);
7252 +
7253 + lower_file = unionfs_lower_file(file);
7254 + BUG_ON(!lower_file);
7255 + /*
7256 + * XXX: vm_ops->fault may be called in parallel. Because we have to
7257 + * resort to temporarily changing the vma->vm_file to point to the
7258 + * lower file, a concurrent invocation of unionfs_fault could see a
7259 + * different value. In this workaround, we keep a different copy of
7260 + * the vma structure in our stack, so we never expose a different
7261 + * value of the vma->vm_file called to us, even temporarily. A
7262 + * better fix would be to change the calling semantics of ->fault to
7263 + * take an explicit file pointer.
7264 + */
7265 + lower_vma.vm_file = lower_file;
7266 + err = lower_vm_ops->fault(&lower_vma, vmf);
7267 + return err;
7268 +}
7269 +
7270 +/*
7271 + * XXX: the default address_space_ops for unionfs is empty. We cannot set
7272 + * our inode->i_mapping->a_ops to NULL because too many code paths expect
7273 + * the a_ops vector to be non-NULL.
7274 + */
7275 +struct address_space_operations unionfs_aops = {
7276 + /* empty on purpose */
7277 +};
7278 +
7279 +/*
7280 + * XXX: we need a second, dummy address_space_ops vector, to be used
7281 + * temporarily during unionfs_mmap, because the latter calls
7282 + * generic_file_mmap, which checks if ->readpage exists, else returns
7283 + * -ENOEXEC.
7284 + */
7285 +struct address_space_operations unionfs_dummy_aops = {
7286 + .readpage = unionfs_readpage,
7287 +};
7288 +
7289 +struct vm_operations_struct unionfs_vm_ops = {
7290 + .fault = unionfs_fault,
7291 +};
7292 diff -Naur linux-2.6.30/fs/unionfs/rdstate.c linux-2.6.30-magellan/fs/unionfs/rdstate.c
7293 --- linux-2.6.30/fs/unionfs/rdstate.c 1970-01-01 01:00:00.000000000 +0100
7294 +++ linux-2.6.30-magellan/fs/unionfs/rdstate.c 2009-06-11 18:39:19.000000000 +0200
7295 @@ -0,0 +1,285 @@
7296 +/*
7297 + * Copyright (c) 2003-2009 Erez Zadok
7298 + * Copyright (c) 2003-2006 Charles P. Wright
7299 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
7300 + * Copyright (c) 2005-2006 Junjiro Okajima
7301 + * Copyright (c) 2005 Arun M. Krishnakumar
7302 + * Copyright (c) 2004-2006 David P. Quigley
7303 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
7304 + * Copyright (c) 2003 Puja Gupta
7305 + * Copyright (c) 2003 Harikesavan Krishnan
7306 + * Copyright (c) 2003-2009 Stony Brook University
7307 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
7308 + *
7309 + * This program is free software; you can redistribute it and/or modify
7310 + * it under the terms of the GNU General Public License version 2 as
7311 + * published by the Free Software Foundation.
7312 + */
7313 +
7314 +#include "union.h"
7315 +
7316 +/* This file contains the routines for maintaining readdir state. */
7317 +
7318 +/*
7319 + * There are two structures here, rdstate which is a hash table
7320 + * of the second structure which is a filldir_node.
7321 + */
7322 +
7323 +/*
7324 + * This is a struct kmem_cache for filldir nodes, because we allocate a lot
7325 + * of them and they shouldn't waste memory. If the node has a small name
7326 + * (as defined by the dentry structure), then we use an inline name to
7327 + * preserve kmalloc space.
7328 + */
7329 +static struct kmem_cache *unionfs_filldir_cachep;
7330 +
7331 +int unionfs_init_filldir_cache(void)
7332 +{
7333 + unionfs_filldir_cachep =
7334 + kmem_cache_create("unionfs_filldir",
7335 + sizeof(struct filldir_node), 0,
7336 + SLAB_RECLAIM_ACCOUNT, NULL);
7337 +
7338 + return (unionfs_filldir_cachep ? 0 : -ENOMEM);
7339 +}
7340 +
7341 +void unionfs_destroy_filldir_cache(void)
7342 +{
7343 + if (unionfs_filldir_cachep)
7344 + kmem_cache_destroy(unionfs_filldir_cachep);
7345 +}
7346 +
7347 +/*
7348 + * This is a tuning parameter that tells us roughly how big to make the
7349 + * hash table in directory entries per page. This isn't perfect, but
7350 + * at least we get a hash table size that shouldn't be too overloaded.
7351 + * The following averages are based on my home directory.
7352 + * 14.44693 Overall
7353 + * 12.29 Single Page Directories
7354 + * 117.93 Multi-page directories
7355 + */
7356 +#define DENTPAGE 4096
7357 +#define DENTPERONEPAGE 12
7358 +#define DENTPERPAGE 118
7359 +#define MINHASHSIZE 1
7360 +static int guesstimate_hash_size(struct inode *inode)
7361 +{
7362 + struct inode *lower_inode;
7363 + int bindex;
7364 + int hashsize = MINHASHSIZE;
7365 +
7366 + if (UNIONFS_I(inode)->hashsize > 0)
7367 + return UNIONFS_I(inode)->hashsize;
7368 +
7369 + for (bindex = ibstart(inode); bindex <= ibend(inode); bindex++) {
7370 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
7371 + if (!lower_inode)
7372 + continue;
7373 +
7374 + if (i_size_read(lower_inode) == DENTPAGE)
7375 + hashsize += DENTPERONEPAGE;
7376 + else
7377 + hashsize += (i_size_read(lower_inode) / DENTPAGE) *
7378 + DENTPERPAGE;
7379 + }
7380 +
7381 + return hashsize;
7382 +}
7383 +
7384 +int init_rdstate(struct file *file)
7385 +{
7386 + BUG_ON(sizeof(loff_t) !=
7387 + (sizeof(unsigned int) + sizeof(unsigned int)));
7388 + BUG_ON(UNIONFS_F(file)->rdstate != NULL);
7389 +
7390 + UNIONFS_F(file)->rdstate = alloc_rdstate(file->f_path.dentry->d_inode,
7391 + fbstart(file));
7392 +
7393 + return (UNIONFS_F(file)->rdstate ? 0 : -ENOMEM);
7394 +}
7395 +
7396 +struct unionfs_dir_state *find_rdstate(struct inode *inode, loff_t fpos)
7397 +{
7398 + struct unionfs_dir_state *rdstate = NULL;
7399 + struct list_head *pos;
7400 +
7401 + spin_lock(&UNIONFS_I(inode)->rdlock);
7402 + list_for_each(pos, &UNIONFS_I(inode)->readdircache) {
7403 + struct unionfs_dir_state *r =
7404 + list_entry(pos, struct unionfs_dir_state, cache);
7405 + if (fpos == rdstate2offset(r)) {
7406 + UNIONFS_I(inode)->rdcount--;
7407 + list_del(&r->cache);
7408 + rdstate = r;
7409 + break;
7410 + }
7411 + }
7412 + spin_unlock(&UNIONFS_I(inode)->rdlock);
7413 + return rdstate;
7414 +}
7415 +
7416 +struct unionfs_dir_state *alloc_rdstate(struct inode *inode, int bindex)
7417 +{
7418 + int i = 0;
7419 + int hashsize;
7420 + unsigned long mallocsize = sizeof(struct unionfs_dir_state);
7421 + struct unionfs_dir_state *rdstate;
7422 +
7423 + hashsize = guesstimate_hash_size(inode);
7424 + mallocsize += hashsize * sizeof(struct list_head);
7425 + mallocsize = __roundup_pow_of_two(mallocsize);
7426 +
7427 + /* This should give us about 500 entries anyway. */
7428 + if (mallocsize > PAGE_SIZE)
7429 + mallocsize = PAGE_SIZE;
7430 +
7431 + hashsize = (mallocsize - sizeof(struct unionfs_dir_state)) /
7432 + sizeof(struct list_head);
7433 +
7434 + rdstate = kmalloc(mallocsize, GFP_KERNEL);
7435 + if (unlikely(!rdstate))
7436 + return NULL;
7437 +
7438 + spin_lock(&UNIONFS_I(inode)->rdlock);
7439 + if (UNIONFS_I(inode)->cookie >= (MAXRDCOOKIE - 1))
7440 + UNIONFS_I(inode)->cookie = 1;
7441 + else
7442 + UNIONFS_I(inode)->cookie++;
7443 +
7444 + rdstate->cookie = UNIONFS_I(inode)->cookie;
7445 + spin_unlock(&UNIONFS_I(inode)->rdlock);
7446 + rdstate->offset = 1;
7447 + rdstate->access = jiffies;
7448 + rdstate->bindex = bindex;
7449 + rdstate->dirpos = 0;
7450 + rdstate->hashentries = 0;
7451 + rdstate->size = hashsize;
7452 + for (i = 0; i < rdstate->size; i++)
7453 + INIT_LIST_HEAD(&rdstate->list[i]);
7454 +
7455 + return rdstate;
7456 +}
7457 +
7458 +static void free_filldir_node(struct filldir_node *node)
7459 +{
7460 + if (node->namelen >= DNAME_INLINE_LEN_MIN)
7461 + kfree(node->name);
7462 + kmem_cache_free(unionfs_filldir_cachep, node);
7463 +}
7464 +
7465 +void free_rdstate(struct unionfs_dir_state *state)
7466 +{
7467 + struct filldir_node *tmp;
7468 + int i;
7469 +
7470 + for (i = 0; i < state->size; i++) {
7471 + struct list_head *head = &(state->list[i]);
7472 + struct list_head *pos, *n;
7473 +
7474 + /* traverse the list and deallocate space */
7475 + list_for_each_safe(pos, n, head) {
7476 + tmp = list_entry(pos, struct filldir_node, file_list);
7477 + list_del(&tmp->file_list);
7478 + free_filldir_node(tmp);
7479 + }
7480 + }
7481 +
7482 + kfree(state);
7483 +}
7484 +
7485 +struct filldir_node *find_filldir_node(struct unionfs_dir_state *rdstate,
7486 + const char *name, int namelen,
7487 + int is_whiteout)
7488 +{
7489 + int index;
7490 + unsigned int hash;
7491 + struct list_head *head;
7492 + struct list_head *pos;
7493 + struct filldir_node *cursor = NULL;
7494 + int found = 0;
7495 +
7496 + BUG_ON(namelen <= 0);
7497 +
7498 + hash = full_name_hash(name, namelen);
7499 + index = hash % rdstate->size;
7500 +
7501 + head = &(rdstate->list[index]);
7502 + list_for_each(pos, head) {
7503 + cursor = list_entry(pos, struct filldir_node, file_list);
7504 +
7505 + if (cursor->namelen == namelen && cursor->hash == hash &&
7506 + !strncmp(cursor->name, name, namelen)) {
7507 + /*
7508 + * a duplicate exists, and hence no need to create
7509 + * entry to the list
7510 + */
7511 + found = 1;
7512 +
7513 + /*
7514 + * if a duplicate is found in this branch, and is
7515 + * not due to the caller looking for an entry to
7516 + * whiteout, then the file system may be corrupted.
7517 + */
7518 + if (unlikely(!is_whiteout &&
7519 + cursor->bindex == rdstate->bindex))
7520 + printk(KERN_ERR "unionfs: filldir: possible "
7521 + "I/O error: a file is duplicated "
7522 + "in the same branch %d: %s\n",
7523 + rdstate->bindex, cursor->name);
7524 + break;
7525 + }
7526 + }
7527 +
7528 + if (!found)
7529 + cursor = NULL;
7530 +
7531 + return cursor;
7532 +}
7533 +
7534 +int add_filldir_node(struct unionfs_dir_state *rdstate, const char *name,
7535 + int namelen, int bindex, int whiteout)
7536 +{
7537 + struct filldir_node *new;
7538 + unsigned int hash;
7539 + int index;
7540 + int err = 0;
7541 + struct list_head *head;
7542 +
7543 + BUG_ON(namelen <= 0);
7544 +
7545 + hash = full_name_hash(name, namelen);
7546 + index = hash % rdstate->size;
7547 + head = &(rdstate->list[index]);
7548 +
7549 + new = kmem_cache_alloc(unionfs_filldir_cachep, GFP_KERNEL);
7550 + if (unlikely(!new)) {
7551 + err = -ENOMEM;
7552 + goto out;
7553 + }
7554 +
7555 + INIT_LIST_HEAD(&new->file_list);
7556 + new->namelen = namelen;
7557 + new->hash = hash;
7558 + new->bindex = bindex;
7559 + new->whiteout = whiteout;
7560 +
7561 + if (namelen < DNAME_INLINE_LEN_MIN) {
7562 + new->name = new->iname;
7563 + } else {
7564 + new->name = kmalloc(namelen + 1, GFP_KERNEL);
7565 + if (unlikely(!new->name)) {
7566 + kmem_cache_free(unionfs_filldir_cachep, new);
7567 + new = NULL;
7568 + goto out;
7569 + }
7570 + }
7571 +
7572 + memcpy(new->name, name, namelen);
7573 + new->name[namelen] = '\0';
7574 +
7575 + rdstate->hashentries++;
7576 +
7577 + list_add(&(new->file_list), head);
7578 +out:
7579 + return err;
7580 +}
7581 diff -Naur linux-2.6.30/fs/unionfs/rename.c linux-2.6.30-magellan/fs/unionfs/rename.c
7582 --- linux-2.6.30/fs/unionfs/rename.c 1970-01-01 01:00:00.000000000 +0100
7583 +++ linux-2.6.30-magellan/fs/unionfs/rename.c 2009-06-11 18:39:19.000000000 +0200
7584 @@ -0,0 +1,520 @@
7585 +/*
7586 + * Copyright (c) 2003-2009 Erez Zadok
7587 + * Copyright (c) 2003-2006 Charles P. Wright
7588 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
7589 + * Copyright (c) 2005-2006 Junjiro Okajima
7590 + * Copyright (c) 2005 Arun M. Krishnakumar
7591 + * Copyright (c) 2004-2006 David P. Quigley
7592 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
7593 + * Copyright (c) 2003 Puja Gupta
7594 + * Copyright (c) 2003 Harikesavan Krishnan
7595 + * Copyright (c) 2003-2009 Stony Brook University
7596 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
7597 + *
7598 + * This program is free software; you can redistribute it and/or modify
7599 + * it under the terms of the GNU General Public License version 2 as
7600 + * published by the Free Software Foundation.
7601 + */
7602 +
7603 +#include "union.h"
7604 +
7605 +/*
7606 + * This is a helper function for rename, used when rename ends up with hosed
7607 + * over dentries and we need to revert.
7608 + */
7609 +static int unionfs_refresh_lower_dentry(struct dentry *dentry,
7610 + struct dentry *parent, int bindex)
7611 +{
7612 + struct dentry *lower_dentry;
7613 + struct dentry *lower_parent;
7614 + int err = 0;
7615 +
7616 + verify_locked(dentry);
7617 +
7618 + lower_parent = unionfs_lower_dentry_idx(parent, bindex);
7619 +
7620 + BUG_ON(!S_ISDIR(lower_parent->d_inode->i_mode));
7621 +
7622 + lower_dentry = lookup_one_len(dentry->d_name.name, lower_parent,
7623 + dentry->d_name.len);
7624 + if (IS_ERR(lower_dentry)) {
7625 + err = PTR_ERR(lower_dentry);
7626 + goto out;
7627 + }
7628 +
7629 + dput(unionfs_lower_dentry_idx(dentry, bindex));
7630 + iput(unionfs_lower_inode_idx(dentry->d_inode, bindex));
7631 + unionfs_set_lower_inode_idx(dentry->d_inode, bindex, NULL);
7632 +
7633 + if (!lower_dentry->d_inode) {
7634 + dput(lower_dentry);
7635 + unionfs_set_lower_dentry_idx(dentry, bindex, NULL);
7636 + } else {
7637 + unionfs_set_lower_dentry_idx(dentry, bindex, lower_dentry);
7638 + unionfs_set_lower_inode_idx(dentry->d_inode, bindex,
7639 + igrab(lower_dentry->d_inode));
7640 + }
7641 +
7642 +out:
7643 + return err;
7644 +}
7645 +
7646 +static int __unionfs_rename(struct inode *old_dir, struct dentry *old_dentry,
7647 + struct dentry *old_parent,
7648 + struct inode *new_dir, struct dentry *new_dentry,
7649 + struct dentry *new_parent,
7650 + int bindex)
7651 +{
7652 + int err = 0;
7653 + struct dentry *lower_old_dentry;
7654 + struct dentry *lower_new_dentry;
7655 + struct dentry *lower_old_dir_dentry;
7656 + struct dentry *lower_new_dir_dentry;
7657 + struct dentry *trap;
7658 +
7659 + lower_new_dentry = unionfs_lower_dentry_idx(new_dentry, bindex);
7660 + lower_old_dentry = unionfs_lower_dentry_idx(old_dentry, bindex);
7661 +
7662 + if (!lower_new_dentry) {
7663 + lower_new_dentry =
7664 + create_parents(new_parent->d_inode,
7665 + new_dentry, new_dentry->d_name.name,
7666 + bindex);
7667 + if (IS_ERR(lower_new_dentry)) {
7668 + err = PTR_ERR(lower_new_dentry);
7669 + if (IS_COPYUP_ERR(err))
7670 + goto out;
7671 + printk(KERN_ERR "unionfs: error creating directory "
7672 + "tree for rename, bindex=%d err=%d\n",
7673 + bindex, err);
7674 + goto out;
7675 + }
7676 + }
7677 +
7678 + /* check for and remove whiteout, if any */
7679 + err = check_unlink_whiteout(new_dentry, lower_new_dentry, bindex);
7680 + if (err > 0) /* ignore if whiteout found and successfully removed */
7681 + err = 0;
7682 + if (err)
7683 + goto out;
7684 +
7685 + /* check of old_dentry branch is writable */
7686 + err = is_robranch_super(old_dentry->d_sb, bindex);
7687 + if (err)
7688 + goto out;
7689 +
7690 + dget(lower_old_dentry);
7691 + dget(lower_new_dentry);
7692 + lower_old_dir_dentry = dget_parent(lower_old_dentry);
7693 + lower_new_dir_dentry = dget_parent(lower_new_dentry);
7694 +
7695 + /* see Documentation/filesystems/unionfs/issues.txt */
7696 + lockdep_off();
7697 + trap = lock_rename(lower_old_dir_dentry, lower_new_dir_dentry);
7698 + /* source should not be ancenstor of target */
7699 + if (trap == lower_old_dentry) {
7700 + err = -EINVAL;
7701 + goto out_err_unlock;
7702 + }
7703 + /* target should not be ancenstor of source */
7704 + if (trap == lower_new_dentry) {
7705 + err = -ENOTEMPTY;
7706 + goto out_err_unlock;
7707 + }
7708 + err = vfs_rename(lower_old_dir_dentry->d_inode, lower_old_dentry,
7709 + lower_new_dir_dentry->d_inode, lower_new_dentry);
7710 +out_err_unlock:
7711 + if (!err) {
7712 + /* update parent dir times */
7713 + fsstack_copy_attr_times(old_dir, lower_old_dir_dentry->d_inode);
7714 + fsstack_copy_attr_times(new_dir, lower_new_dir_dentry->d_inode);
7715 + }
7716 + unlock_rename(lower_old_dir_dentry, lower_new_dir_dentry);
7717 + lockdep_on();
7718 +
7719 + dput(lower_old_dir_dentry);
7720 + dput(lower_new_dir_dentry);
7721 + dput(lower_old_dentry);
7722 + dput(lower_new_dentry);
7723 +
7724 +out:
7725 + if (!err) {
7726 + /* Fixup the new_dentry. */
7727 + if (bindex < dbstart(new_dentry))
7728 + dbstart(new_dentry) = bindex;
7729 + else if (bindex > dbend(new_dentry))
7730 + dbend(new_dentry) = bindex;
7731 + }
7732 +
7733 + return err;
7734 +}
7735 +
7736 +/*
7737 + * Main rename code. This is sufficiently complex, that it's documented in
7738 + * Documentation/filesystems/unionfs/rename.txt. This routine calls
7739 + * __unionfs_rename() above to perform some of the work.
7740 + */
7741 +static int do_unionfs_rename(struct inode *old_dir,
7742 + struct dentry *old_dentry,
7743 + struct dentry *old_parent,
7744 + struct inode *new_dir,
7745 + struct dentry *new_dentry,
7746 + struct dentry *new_parent)
7747 +{
7748 + int err = 0;
7749 + int bindex;
7750 + int old_bstart, old_bend;
7751 + int new_bstart, new_bend;
7752 + int do_copyup = -1;
7753 + int local_err = 0;
7754 + int eio = 0;
7755 + int revert = 0;
7756 +
7757 + old_bstart = dbstart(old_dentry);
7758 + old_bend = dbend(old_dentry);
7759 +
7760 + new_bstart = dbstart(new_dentry);
7761 + new_bend = dbend(new_dentry);
7762 +
7763 + /* Rename source to destination. */
7764 + err = __unionfs_rename(old_dir, old_dentry, old_parent,
7765 + new_dir, new_dentry, new_parent,
7766 + old_bstart);
7767 + if (err) {
7768 + if (!IS_COPYUP_ERR(err))
7769 + goto out;
7770 + do_copyup = old_bstart - 1;
7771 + } else {
7772 + revert = 1;
7773 + }
7774 +
7775 + /*
7776 + * Unlink all instances of destination that exist to the left of
7777 + * bstart of source. On error, revert back, goto out.
7778 + */
7779 + for (bindex = old_bstart - 1; bindex >= new_bstart; bindex--) {
7780 + struct dentry *unlink_dentry;
7781 + struct dentry *unlink_dir_dentry;
7782 +
7783 + BUG_ON(bindex < 0);
7784 + unlink_dentry = unionfs_lower_dentry_idx(new_dentry, bindex);
7785 + if (!unlink_dentry)
7786 + continue;
7787 +
7788 + unlink_dir_dentry = lock_parent(unlink_dentry);
7789 + err = is_robranch_super(old_dir->i_sb, bindex);
7790 + if (!err)
7791 + err = vfs_unlink(unlink_dir_dentry->d_inode,
7792 + unlink_dentry);
7793 +
7794 + fsstack_copy_attr_times(new_parent->d_inode,
7795 + unlink_dir_dentry->d_inode);
7796 + /* propagate number of hard-links */
7797 + new_parent->d_inode->i_nlink =
7798 + unionfs_get_nlinks(new_parent->d_inode);
7799 +
7800 + unlock_dir(unlink_dir_dentry);
7801 + if (!err) {
7802 + if (bindex != new_bstart) {
7803 + dput(unlink_dentry);
7804 + unionfs_set_lower_dentry_idx(new_dentry,
7805 + bindex, NULL);
7806 + }
7807 + } else if (IS_COPYUP_ERR(err)) {
7808 + do_copyup = bindex - 1;
7809 + } else if (revert) {
7810 + goto revert;
7811 + }
7812 + }
7813 +
7814 + if (do_copyup != -1) {
7815 + for (bindex = do_copyup; bindex >= 0; bindex--) {
7816 + /*
7817 + * copyup the file into some left directory, so that
7818 + * you can rename it
7819 + */
7820 + err = copyup_dentry(old_parent->d_inode,
7821 + old_dentry, old_bstart, bindex,
7822 + old_dentry->d_name.name,
7823 + old_dentry->d_name.len, NULL,
7824 + i_size_read(old_dentry->d_inode));
7825 + /* if copyup failed, try next branch to the left */
7826 + if (err)
7827 + continue;
7828 + /*
7829 + * create whiteout before calling __unionfs_rename
7830 + * because the latter will change the old_dentry's
7831 + * lower name and parent dir, resulting in the
7832 + * whiteout getting created in the wrong dir.
7833 + */
7834 + err = create_whiteout(old_dentry, bindex);
7835 + if (err) {
7836 + printk(KERN_ERR "unionfs: can't create a "
7837 + "whiteout for %s in rename (err=%d)\n",
7838 + old_dentry->d_name.name, err);
7839 + continue;
7840 + }
7841 + err = __unionfs_rename(old_dir, old_dentry, old_parent,
7842 + new_dir, new_dentry, new_parent,
7843 + bindex);
7844 + break;
7845 + }
7846 + }
7847 +
7848 + /* make it opaque */
7849 + if (S_ISDIR(old_dentry->d_inode->i_mode)) {
7850 + err = make_dir_opaque(old_dentry, dbstart(old_dentry));
7851 + if (err)
7852 + goto revert;
7853 + }
7854 +
7855 + /*
7856 + * Create whiteout for source, only if:
7857 + * (1) There is more than one underlying instance of source.
7858 + * (We did a copy_up is taken care of above).
7859 + */
7860 + if ((old_bstart != old_bend) && (do_copyup == -1)) {
7861 + err = create_whiteout(old_dentry, old_bstart);
7862 + if (err) {
7863 + /* can't fix anything now, so we exit with -EIO */
7864 + printk(KERN_ERR "unionfs: can't create a whiteout for "
7865 + "%s in rename!\n", old_dentry->d_name.name);
7866 + err = -EIO;
7867 + }
7868 + }
7869 +
7870 +out:
7871 + return err;
7872 +
7873 +revert:
7874 + /* Do revert here. */
7875 + local_err = unionfs_refresh_lower_dentry(new_dentry, new_parent,
7876 + old_bstart);
7877 + if (local_err) {
7878 + printk(KERN_ERR "unionfs: revert failed in rename: "
7879 + "the new refresh failed\n");
7880 + eio = -EIO;
7881 + }
7882 +
7883 + local_err = unionfs_refresh_lower_dentry(old_dentry, old_parent,
7884 + old_bstart);
7885 + if (local_err) {
7886 + printk(KERN_ERR "unionfs: revert failed in rename: "
7887 + "the old refresh failed\n");
7888 + eio = -EIO;
7889 + goto revert_out;
7890 + }
7891 +
7892 + if (!unionfs_lower_dentry_idx(new_dentry, bindex) ||
7893 + !unionfs_lower_dentry_idx(new_dentry, bindex)->d_inode) {
7894 + printk(KERN_ERR "unionfs: revert failed in rename: "
7895 + "the object disappeared from under us!\n");
7896 + eio = -EIO;
7897 + goto revert_out;
7898 + }
7899 +
7900 + if (unionfs_lower_dentry_idx(old_dentry, bindex) &&
7901 + unionfs_lower_dentry_idx(old_dentry, bindex)->d_inode) {
7902 + printk(KERN_ERR "unionfs: revert failed in rename: "
7903 + "the object was created underneath us!\n");
7904 + eio = -EIO;
7905 + goto revert_out;
7906 + }
7907 +
7908 + local_err = __unionfs_rename(new_dir, new_dentry, new_parent,
7909 + old_dir, old_dentry, old_parent,
7910 + old_bstart);
7911 +
7912 + /* If we can't fix it, then we cop-out with -EIO. */
7913 + if (local_err) {
7914 + printk(KERN_ERR "unionfs: revert failed in rename!\n");
7915 + eio = -EIO;
7916 + }
7917 +
7918 + local_err = unionfs_refresh_lower_dentry(new_dentry, new_parent,
7919 + bindex);
7920 + if (local_err)
7921 + eio = -EIO;
7922 + local_err = unionfs_refresh_lower_dentry(old_dentry, old_parent,
7923 + bindex);
7924 + if (local_err)
7925 + eio = -EIO;
7926 +
7927 +revert_out:
7928 + if (eio)
7929 + err = eio;
7930 + return err;
7931 +}
7932 +
7933 +/*
7934 + * We can't copyup a directory, because it may involve huge numbers of
7935 + * children, etc. Doing that in the kernel would be bad, so instead we
7936 + * return EXDEV to the user-space utility that caused this, and let the
7937 + * user-space recurse and ask us to copy up each file separately.
7938 + */
7939 +static int may_rename_dir(struct dentry *dentry, struct dentry *parent)
7940 +{
7941 + int err, bstart;
7942 +
7943 + err = check_empty(dentry, parent, NULL);
7944 + if (err == -ENOTEMPTY) {
7945 + if (is_robranch(dentry))
7946 + return -EXDEV;
7947 + } else if (err) {
7948 + return err;
7949 + }
7950 +
7951 + bstart = dbstart(dentry);
7952 + if (dbend(dentry) == bstart || dbopaque(dentry) == bstart)
7953 + return 0;
7954 +
7955 + dbstart(dentry) = bstart + 1;
7956 + err = check_empty(dentry, parent, NULL);
7957 + dbstart(dentry) = bstart;
7958 + if (err == -ENOTEMPTY)
7959 + err = -EXDEV;
7960 + return err;
7961 +}
7962 +
7963 +/*
7964 + * The locking rules in unionfs_rename are complex. We could use a simpler
7965 + * superblock-level name-space lock for renames and copy-ups.
7966 + */
7967 +int unionfs_rename(struct inode *old_dir, struct dentry *old_dentry,
7968 + struct inode *new_dir, struct dentry *new_dentry)
7969 +{
7970 + int err = 0;
7971 + struct dentry *wh_dentry;
7972 + struct dentry *old_parent, *new_parent;
7973 + int valid = true;
7974 +
7975 + unionfs_read_lock(old_dentry->d_sb, UNIONFS_SMUTEX_CHILD);
7976 + old_parent = dget_parent(old_dentry);
7977 + new_parent = dget_parent(new_dentry);
7978 + /* un/lock parent dentries only if they differ from old/new_dentry */
7979 + if (old_parent != old_dentry &&
7980 + old_parent != new_dentry)
7981 + unionfs_lock_dentry(old_parent, UNIONFS_DMUTEX_REVAL_PARENT);
7982 + if (new_parent != old_dentry &&
7983 + new_parent != new_dentry &&
7984 + new_parent != old_parent)
7985 + unionfs_lock_dentry(new_parent, UNIONFS_DMUTEX_REVAL_CHILD);
7986 + unionfs_double_lock_dentry(old_dentry, new_dentry);
7987 +
7988 + valid = __unionfs_d_revalidate(old_dentry, old_parent, false);
7989 + if (!valid) {
7990 + err = -ESTALE;
7991 + goto out;
7992 + }
7993 + if (!d_deleted(new_dentry) && new_dentry->d_inode) {
7994 + valid = __unionfs_d_revalidate(new_dentry, new_parent, false);
7995 + if (!valid) {
7996 + err = -ESTALE;
7997 + goto out;
7998 + }
7999 + }
8000 +
8001 + if (!S_ISDIR(old_dentry->d_inode->i_mode))
8002 + err = unionfs_partial_lookup(old_dentry, old_parent);
8003 + else
8004 + err = may_rename_dir(old_dentry, old_parent);
8005 +
8006 + if (err)
8007 + goto out;
8008 +
8009 + err = unionfs_partial_lookup(new_dentry, new_parent);
8010 + if (err)
8011 + goto out;
8012 +
8013 + /*
8014 + * if new_dentry is already lower because of whiteout,
8015 + * simply override it even if the whited-out dir is not empty.
8016 + */
8017 + wh_dentry = find_first_whiteout(new_dentry);
8018 + if (!IS_ERR(wh_dentry)) {
8019 + dput(wh_dentry);
8020 + } else if (new_dentry->d_inode) {
8021 + if (S_ISDIR(old_dentry->d_inode->i_mode) !=
8022 + S_ISDIR(new_dentry->d_inode->i_mode)) {
8023 + err = S_ISDIR(old_dentry->d_inode->i_mode) ?
8024 + -ENOTDIR : -EISDIR;
8025 + goto out;
8026 + }
8027 +
8028 + if (S_ISDIR(new_dentry->d_inode->i_mode)) {
8029 + struct unionfs_dir_state *namelist = NULL;
8030 + /* check if this unionfs directory is empty or not */
8031 + err = check_empty(new_dentry, new_parent, &namelist);
8032 + if (err)
8033 + goto out;
8034 +
8035 + if (!is_robranch(new_dentry))
8036 + err = delete_whiteouts(new_dentry,
8037 + dbstart(new_dentry),
8038 + namelist);
8039 +
8040 + free_rdstate(namelist);
8041 +
8042 + if (err)
8043 + goto out;
8044 + }
8045 + }
8046 +
8047 + err = do_unionfs_rename(old_dir, old_dentry, old_parent,
8048 + new_dir, new_dentry, new_parent);
8049 + if (err)
8050 + goto out;
8051 +
8052 + /*
8053 + * force re-lookup since the dir on ro branch is not renamed, and
8054 + * lower dentries still indicate the un-renamed ones.
8055 + */
8056 + if (S_ISDIR(old_dentry->d_inode->i_mode))
8057 + atomic_dec(&UNIONFS_D(old_dentry)->generation);
8058 + else
8059 + unionfs_postcopyup_release(old_dentry);
8060 + if (new_dentry->d_inode && !S_ISDIR(new_dentry->d_inode->i_mode)) {
8061 + unionfs_postcopyup_release(new_dentry);
8062 + unionfs_postcopyup_setmnt(new_dentry);
8063 + if (!unionfs_lower_inode(new_dentry->d_inode)) {
8064 + /*
8065 + * If we get here, it means that no copyup was
8066 + * needed, and that a file by the old name already
8067 + * existing on the destination branch; that file got
8068 + * renamed earlier in this function, so all we need
8069 + * to do here is set the lower inode.
8070 + */
8071 + struct inode *inode;
8072 + inode = unionfs_lower_inode(old_dentry->d_inode);
8073 + igrab(inode);
8074 + unionfs_set_lower_inode_idx(new_dentry->d_inode,
8075 + dbstart(new_dentry),
8076 + inode);
8077 + }
8078 + }
8079 + /* if all of this renaming succeeded, update our times */
8080 + unionfs_copy_attr_times(old_dentry->d_inode);
8081 + unionfs_copy_attr_times(new_dentry->d_inode);
8082 + unionfs_check_inode(old_dir);
8083 + unionfs_check_inode(new_dir);
8084 + unionfs_check_dentry(old_dentry);
8085 + unionfs_check_dentry(new_dentry);
8086 +
8087 +out:
8088 + if (err) /* clear the new_dentry stuff created */
8089 + d_drop(new_dentry);
8090 +
8091 + unionfs_double_unlock_dentry(old_dentry, new_dentry);
8092 + if (new_parent != old_dentry &&
8093 + new_parent != new_dentry &&
8094 + new_parent != old_parent)
8095 + unionfs_unlock_dentry(new_parent);
8096 + if (old_parent != old_dentry &&
8097 + old_parent != new_dentry)
8098 + unionfs_unlock_dentry(old_parent);
8099 + dput(new_parent);
8100 + dput(old_parent);
8101 + unionfs_read_unlock(old_dentry->d_sb);
8102 +
8103 + return err;
8104 +}
8105 diff -Naur linux-2.6.30/fs/unionfs/sioq.c linux-2.6.30-magellan/fs/unionfs/sioq.c
8106 --- linux-2.6.30/fs/unionfs/sioq.c 1970-01-01 01:00:00.000000000 +0100
8107 +++ linux-2.6.30-magellan/fs/unionfs/sioq.c 2009-06-11 18:39:19.000000000 +0200
8108 @@ -0,0 +1,101 @@
8109 +/*
8110 + * Copyright (c) 2006-2009 Erez Zadok
8111 + * Copyright (c) 2006 Charles P. Wright
8112 + * Copyright (c) 2006-2007 Josef 'Jeff' Sipek
8113 + * Copyright (c) 2006 Junjiro Okajima
8114 + * Copyright (c) 2006 David P. Quigley
8115 + * Copyright (c) 2006-2009 Stony Brook University
8116 + * Copyright (c) 2006-2009 The Research Foundation of SUNY
8117 + *
8118 + * This program is free software; you can redistribute it and/or modify
8119 + * it under the terms of the GNU General Public License version 2 as
8120 + * published by the Free Software Foundation.
8121 + */
8122 +
8123 +#include "union.h"
8124 +
8125 +/*
8126 + * Super-user IO work Queue - sometimes we need to perform actions which
8127 + * would fail due to the unix permissions on the parent directory (e.g.,
8128 + * rmdir a directory which appears empty, but in reality contains
8129 + * whiteouts).
8130 + */
8131 +
8132 +static struct workqueue_struct *superio_workqueue;
8133 +
8134 +int __init init_sioq(void)
8135 +{
8136 + int err;
8137 +
8138 + superio_workqueue = create_workqueue("unionfs_siod");
8139 + if (!IS_ERR(superio_workqueue))
8140 + return 0;
8141 +
8142 + err = PTR_ERR(superio_workqueue);
8143 + printk(KERN_ERR "unionfs: create_workqueue failed %d\n", err);
8144 + superio_workqueue = NULL;
8145 + return err;
8146 +}
8147 +
8148 +void stop_sioq(void)
8149 +{
8150 + if (superio_workqueue)
8151 + destroy_workqueue(superio_workqueue);
8152 +}
8153 +
8154 +void run_sioq(work_func_t func, struct sioq_args *args)
8155 +{
8156 + INIT_WORK(&args->work, func);
8157 +
8158 + init_completion(&args->comp);
8159 + while (!queue_work(superio_workqueue, &args->work)) {
8160 + /* TODO: do accounting if needed */
8161 + schedule();
8162 + }
8163 + wait_for_completion(&args->comp);
8164 +}
8165 +
8166 +void __unionfs_create(struct work_struct *work)
8167 +{
8168 + struct sioq_args *args = container_of(work, struct sioq_args, work);
8169 + struct create_args *c = &args->create;
8170 +
8171 + args->err = vfs_create(c->parent, c->dentry, c->mode, c->nd);
8172 + complete(&args->comp);
8173 +}
8174 +
8175 +void __unionfs_mkdir(struct work_struct *work)
8176 +{
8177 + struct sioq_args *args = container_of(work, struct sioq_args, work);
8178 + struct mkdir_args *m = &args->mkdir;
8179 +
8180 + args->err = vfs_mkdir(m->parent, m->dentry, m->mode);
8181 + complete(&args->comp);
8182 +}
8183 +
8184 +void __unionfs_mknod(struct work_struct *work)
8185 +{
8186 + struct sioq_args *args = container_of(work, struct sioq_args, work);
8187 + struct mknod_args *m = &args->mknod;
8188 +
8189 + args->err = vfs_mknod(m->parent, m->dentry, m->mode, m->dev);
8190 + complete(&args->comp);
8191 +}
8192 +
8193 +void __unionfs_symlink(struct work_struct *work)
8194 +{
8195 + struct sioq_args *args = container_of(work, struct sioq_args, work);
8196 + struct symlink_args *s = &args->symlink;
8197 +
8198 + args->err = vfs_symlink(s->parent, s->dentry, s->symbuf);
8199 + complete(&args->comp);
8200 +}
8201 +
8202 +void __unionfs_unlink(struct work_struct *work)
8203 +{
8204 + struct sioq_args *args = container_of(work, struct sioq_args, work);
8205 + struct unlink_args *u = &args->unlink;
8206 +
8207 + args->err = vfs_unlink(u->parent, u->dentry);
8208 + complete(&args->comp);
8209 +}
8210 diff -Naur linux-2.6.30/fs/unionfs/sioq.h linux-2.6.30-magellan/fs/unionfs/sioq.h
8211 --- linux-2.6.30/fs/unionfs/sioq.h 1970-01-01 01:00:00.000000000 +0100
8212 +++ linux-2.6.30-magellan/fs/unionfs/sioq.h 2009-06-11 18:39:19.000000000 +0200
8213 @@ -0,0 +1,91 @@
8214 +/*
8215 + * Copyright (c) 2006-2009 Erez Zadok
8216 + * Copyright (c) 2006 Charles P. Wright
8217 + * Copyright (c) 2006-2007 Josef 'Jeff' Sipek
8218 + * Copyright (c) 2006 Junjiro Okajima
8219 + * Copyright (c) 2006 David P. Quigley
8220 + * Copyright (c) 2006-2009 Stony Brook University
8221 + * Copyright (c) 2006-2009 The Research Foundation of SUNY
8222 + *
8223 + * This program is free software; you can redistribute it and/or modify
8224 + * it under the terms of the GNU General Public License version 2 as
8225 + * published by the Free Software Foundation.
8226 + */
8227 +
8228 +#ifndef _SIOQ_H
8229 +#define _SIOQ_H
8230 +
8231 +struct deletewh_args {
8232 + struct unionfs_dir_state *namelist;
8233 + struct dentry *dentry;
8234 + int bindex;
8235 +};
8236 +
8237 +struct is_opaque_args {
8238 + struct dentry *dentry;
8239 +};
8240 +
8241 +struct create_args {
8242 + struct inode *parent;
8243 + struct dentry *dentry;
8244 + umode_t mode;
8245 + struct nameidata *nd;
8246 +};
8247 +
8248 +struct mkdir_args {
8249 + struct inode *parent;
8250 + struct dentry *dentry;
8251 + umode_t mode;
8252 +};
8253 +
8254 +struct mknod_args {
8255 + struct inode *parent;
8256 + struct dentry *dentry;
8257 + umode_t mode;
8258 + dev_t dev;
8259 +};
8260 +
8261 +struct symlink_args {
8262 + struct inode *parent;
8263 + struct dentry *dentry;
8264 + char *symbuf;
8265 +};
8266 +
8267 +struct unlink_args {
8268 + struct inode *parent;
8269 + struct dentry *dentry;
8270 +};
8271 +
8272 +
8273 +struct sioq_args {
8274 + struct completion comp;
8275 + struct work_struct work;
8276 + int err;
8277 + void *ret;
8278 +
8279 + union {
8280 + struct deletewh_args deletewh;
8281 + struct is_opaque_args is_opaque;
8282 + struct create_args create;
8283 + struct mkdir_args mkdir;
8284 + struct mknod_args mknod;
8285 + struct symlink_args symlink;
8286 + struct unlink_args unlink;
8287 + };
8288 +};
8289 +
8290 +/* Extern definitions for SIOQ functions */
8291 +extern int __init init_sioq(void);
8292 +extern void stop_sioq(void);
8293 +extern void run_sioq(work_func_t func, struct sioq_args *args);
8294 +
8295 +/* Extern definitions for our privilege escalation helpers */
8296 +extern void __unionfs_create(struct work_struct *work);
8297 +extern void __unionfs_mkdir(struct work_struct *work);
8298 +extern void __unionfs_mknod(struct work_struct *work);
8299 +extern void __unionfs_symlink(struct work_struct *work);
8300 +extern void __unionfs_unlink(struct work_struct *work);
8301 +extern void __delete_whiteouts(struct work_struct *work);
8302 +extern void __is_opaque_dir(struct work_struct *work);
8303 +
8304 +#endif /* not _SIOQ_H */
8305 diff -Naur linux-2.6.30/fs/unionfs/subr.c linux-2.6.30-magellan/fs/unionfs/subr.c
8306 --- linux-2.6.30/fs/unionfs/subr.c 1970-01-01 01:00:00.000000000 +0100
8307 +++ linux-2.6.30-magellan/fs/unionfs/subr.c 2009-06-11 18:39:19.000000000 +0200
8308 @@ -0,0 +1,95 @@
8309 +/*
8310 + * Copyright (c) 2003-2009 Erez Zadok
8311 + * Copyright (c) 2003-2006 Charles P. Wright
8312 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
8313 + * Copyright (c) 2005-2006 Junjiro Okajima
8314 + * Copyright (c) 2005 Arun M. Krishnakumar
8315 + * Copyright (c) 2004-2006 David P. Quigley
8316 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
8317 + * Copyright (c) 2003 Puja Gupta
8318 + * Copyright (c) 2003 Harikesavan Krishnan
8319 + * Copyright (c) 2003-2009 Stony Brook University
8320 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
8321 + *
8322 + * This program is free software; you can redistribute it and/or modify
8323 + * it under the terms of the GNU General Public License version 2 as
8324 + * published by the Free Software Foundation.
8325 + */
8326 +
8327 +#include "union.h"
8328 +
8329 +/*
8330 + * returns the right n_link value based on the inode type
8331 + */
8332 +int unionfs_get_nlinks(const struct inode *inode)
8333 +{
8334 + /* don't bother to do all the work since we're unlinked */
8335 + if (inode->i_nlink == 0)
8336 + return 0;
8337 +
8338 + if (!S_ISDIR(inode->i_mode))
8339 + return unionfs_lower_inode(inode)->i_nlink;
8340 +
8341 + /*
8342 + * For directories, we return 1. The only place that could cares
8343 + * about links is readdir, and there's d_type there so even that
8344 + * doesn't matter.
8345 + */
8346 + return 1;
8347 +}
8348 +
8349 +/* copy a/m/ctime from the lower branch with the newest times */
8350 +void unionfs_copy_attr_times(struct inode *upper)
8351 +{
8352 + int bindex;
8353 + struct inode *lower;
8354 +
8355 + if (!upper)
8356 + return;
8357 + if (ibstart(upper) < 0) {
8358 +#ifdef CONFIG_UNION_FS_DEBUG
8359 + WARN_ON(ibstart(upper) < 0);
8360 +#endif /* CONFIG_UNION_FS_DEBUG */
8361 + return;
8362 + }
8363 + for (bindex = ibstart(upper); bindex <= ibend(upper); bindex++) {
8364 + lower = unionfs_lower_inode_idx(upper, bindex);
8365 + if (!lower)
8366 + continue; /* not all lower dir objects may exist */
8367 + if (unlikely(timespec_compare(&upper->i_mtime,
8368 + &lower->i_mtime) < 0))
8369 + upper->i_mtime = lower->i_mtime;
8370 + if (unlikely(timespec_compare(&upper->i_ctime,
8371 + &lower->i_ctime) < 0))
8372 + upper->i_ctime = lower->i_ctime;
8373 + if (unlikely(timespec_compare(&upper->i_atime,
8374 + &lower->i_atime) < 0))
8375 + upper->i_atime = lower->i_atime;
8376 + }
8377 +}
8378 +
8379 +/*
8380 + * A unionfs/fanout version of fsstack_copy_attr_all. Uses a
8381 + * unionfs_get_nlinks to properly calcluate the number of links to a file.
8382 + * Also, copies the max() of all a/m/ctimes for all lower inodes (which is
8383 + * important if the lower inode is a directory type)
8384 + */
8385 +void unionfs_copy_attr_all(struct inode *dest,
8386 + const struct inode *src)
8387 +{
8388 + dest->i_mode = src->i_mode;
8389 + dest->i_uid = src->i_uid;
8390 + dest->i_gid = src->i_gid;
8391 + dest->i_rdev = src->i_rdev;
8392 +
8393 + unionfs_copy_attr_times(dest);
8394 +
8395 + dest->i_blkbits = src->i_blkbits;
8396 + dest->i_flags = src->i_flags;
8397 +
8398 + /*
8399 + * Update the nlinks AFTER updating the above fields, because the
8400 + * get_links callback may depend on them.
8401 + */
8402 + dest->i_nlink = unionfs_get_nlinks(dest);
8403 +}
8404 diff -Naur linux-2.6.30/fs/unionfs/super.c linux-2.6.30-magellan/fs/unionfs/super.c
8405 --- linux-2.6.30/fs/unionfs/super.c 1970-01-01 01:00:00.000000000 +0100
8406 +++ linux-2.6.30-magellan/fs/unionfs/super.c 2009-06-11 18:39:19.000000000 +0200
8407 @@ -0,0 +1,1047 @@
8408 +/*
8409 + * Copyright (c) 2003-2009 Erez Zadok
8410 + * Copyright (c) 2003-2006 Charles P. Wright
8411 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
8412 + * Copyright (c) 2005-2006 Junjiro Okajima
8413 + * Copyright (c) 2005 Arun M. Krishnakumar
8414 + * Copyright (c) 2004-2006 David P. Quigley
8415 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
8416 + * Copyright (c) 2003 Puja Gupta
8417 + * Copyright (c) 2003 Harikesavan Krishnan
8418 + * Copyright (c) 2003-2009 Stony Brook University
8419 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
8420 + *
8421 + * This program is free software; you can redistribute it and/or modify
8422 + * it under the terms of the GNU General Public License version 2 as
8423 + * published by the Free Software Foundation.
8424 + */
8425 +
8426 +#include "union.h"
8427 +
8428 +/*
8429 + * The inode cache is used with alloc_inode for both our inode info and the
8430 + * vfs inode.
8431 + */
8432 +static struct kmem_cache *unionfs_inode_cachep;
8433 +
8434 +struct inode *unionfs_iget(struct super_block *sb, unsigned long ino)
8435 +{
8436 + int size;
8437 + struct unionfs_inode_info *info;
8438 + struct inode *inode;
8439 +
8440 + inode = iget_locked(sb, ino);
8441 + if (!inode)
8442 + return ERR_PTR(-ENOMEM);
8443 + if (!(inode->i_state & I_NEW))
8444 + return inode;
8445 +
8446 + info = UNIONFS_I(inode);
8447 + memset(info, 0, offsetof(struct unionfs_inode_info, vfs_inode));
8448 + info->bstart = -1;
8449 + info->bend = -1;
8450 + atomic_set(&info->generation,
8451 + atomic_read(&UNIONFS_SB(inode->i_sb)->generation));
8452 + spin_lock_init(&info->rdlock);
8453 + info->rdcount = 1;
8454 + info->hashsize = -1;
8455 + INIT_LIST_HEAD(&info->readdircache);
8456 +
8457 + size = sbmax(inode->i_sb) * sizeof(struct inode *);
8458 + info->lower_inodes = kzalloc(size, GFP_KERNEL);
8459 + if (unlikely(!info->lower_inodes)) {
8460 + printk(KERN_CRIT "unionfs: no kernel memory when allocating "
8461 + "lower-pointer array!\n");
8462 + iget_failed(inode);
8463 + return ERR_PTR(-ENOMEM);
8464 + }
8465 +
8466 + inode->i_version++;
8467 + inode->i_op = &unionfs_main_iops;
8468 + inode->i_fop = &unionfs_main_fops;
8469 +
8470 + inode->i_mapping->a_ops = &unionfs_aops;
8471 +
8472 + /*
8473 + * reset times so unionfs_copy_attr_all can keep out time invariants
8474 + * right (upper inode time being the max of all lower ones).
8475 + */
8476 + inode->i_atime.tv_sec = inode->i_atime.tv_nsec = 0;
8477 + inode->i_mtime.tv_sec = inode->i_mtime.tv_nsec = 0;
8478 + inode->i_ctime.tv_sec = inode->i_ctime.tv_nsec = 0;
8479 + unlock_new_inode(inode);
8480 + return inode;
8481 +}
8482 +
8483 +/*
8484 + * we now define delete_inode, because there are two VFS paths that may
8485 + * destroy an inode: one of them calls clear inode before doing everything
8486 + * else that's needed, and the other is fine. This way we truncate the inode
8487 + * size (and its pages) and then clear our own inode, which will do an iput
8488 + * on our and the lower inode.
8489 + *
8490 + * No need to lock sb info's rwsem.
8491 + */
8492 +static void unionfs_delete_inode(struct inode *inode)
8493 +{
8494 +#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
8495 + spin_lock(&inode->i_lock);
8496 +#endif
8497 + i_size_write(inode, 0); /* every f/s seems to do that */
8498 +#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
8499 + spin_unlock(&inode->i_lock);
8500 +#endif
8501 +
8502 + if (inode->i_data.nrpages)
8503 + truncate_inode_pages(&inode->i_data, 0);
8504 +
8505 + clear_inode(inode);
8506 +}
8507 +
8508 +/*
8509 + * final actions when unmounting a file system
8510 + *
8511 + * No need to lock rwsem.
8512 + */
8513 +static void unionfs_put_super(struct super_block *sb)
8514 +{
8515 + int bindex, bstart, bend;
8516 + struct unionfs_sb_info *spd;
8517 + int leaks = 0;
8518 +
8519 + spd = UNIONFS_SB(sb);
8520 + if (!spd)
8521 + return;
8522 +
8523 + bstart = sbstart(sb);
8524 + bend = sbend(sb);
8525 +
8526 + /* Make sure we have no leaks of branchget/branchput. */
8527 + for (bindex = bstart; bindex <= bend; bindex++)
8528 + if (unlikely(branch_count(sb, bindex) != 0)) {
8529 + printk(KERN_CRIT
8530 + "unionfs: branch %d has %d references left!\n",
8531 + bindex, branch_count(sb, bindex));
8532 + leaks = 1;
8533 + }
8534 + WARN_ON(leaks != 0);
8535 +
8536 + /* decrement lower super references */
8537 + for (bindex = bstart; bindex <= bend; bindex++) {
8538 + struct super_block *s;
8539 + s = unionfs_lower_super_idx(sb, bindex);
8540 + unionfs_set_lower_super_idx(sb, bindex, NULL);
8541 + atomic_dec(&s->s_active);
8542 + }
8543 +
8544 + kfree(spd->dev_name);
8545 + kfree(spd->data);
8546 + kfree(spd);
8547 + sb->s_fs_info = NULL;
8548 +}
8549 +
8550 +/*
8551 + * Since people use this to answer the "How big of a file can I write?"
8552 + * question, we report the size of the highest priority branch as the size of
8553 + * the union.
8554 + */
8555 +static int unionfs_statfs(struct dentry *dentry, struct kstatfs *buf)
8556 +{
8557 + int err = 0;
8558 + struct super_block *sb;
8559 + struct dentry *lower_dentry;
8560 + struct dentry *parent;
8561 + bool valid;
8562 +
8563 + sb = dentry->d_sb;
8564 +
8565 + unionfs_read_lock(sb, UNIONFS_SMUTEX_CHILD);
8566 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
8567 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
8568 +
8569 + valid = __unionfs_d_revalidate(dentry, parent, false);
8570 + if (unlikely(!valid)) {
8571 + err = -ESTALE;
8572 + goto out;
8573 + }
8574 + unionfs_check_dentry(dentry);
8575 +
8576 + lower_dentry = unionfs_lower_dentry(sb->s_root);
8577 + err = vfs_statfs(lower_dentry, buf);
8578 +
8579 + /* set return buf to our f/s to avoid confusing user-level utils */
8580 + buf->f_type = UNIONFS_SUPER_MAGIC;
8581 + /*
8582 + * Our maximum file name can is shorter by a few bytes because every
8583 + * file name could potentially be whited-out.
8584 + *
8585 + * XXX: this restriction goes away with ODF.
8586 + */
8587 + unionfs_set_max_namelen(&buf->f_namelen);
8588 +
8589 + /*
8590 + * reset two fields to avoid confusing user-land.
8591 + * XXX: is this still necessary?
8592 + */
8593 + memset(&buf->f_fsid, 0, sizeof(__kernel_fsid_t));
8594 + memset(&buf->f_spare, 0, sizeof(buf->f_spare));
8595 +
8596 +out:
8597 + unionfs_check_dentry(dentry);
8598 + unionfs_unlock_dentry(dentry);
8599 + unionfs_unlock_parent(dentry, parent);
8600 + unionfs_read_unlock(sb);
8601 + return err;
8602 +}
8603 +
8604 +/* handle mode changing during remount */
8605 +static noinline_for_stack int do_remount_mode_option(
8606 + char *optarg,
8607 + int cur_branches,
8608 + struct unionfs_data *new_data,
8609 + struct path *new_lower_paths)
8610 +{
8611 + int err = -EINVAL;
8612 + int perms, idx;
8613 + char *modename = strchr(optarg, '=');
8614 + struct nameidata nd;
8615 +
8616 + /* by now, optarg contains the branch name */
8617 + if (!*optarg) {
8618 + printk(KERN_ERR
8619 + "unionfs: no branch specified for mode change\n");
8620 + goto out;
8621 + }
8622 + if (!modename) {
8623 + printk(KERN_ERR "unionfs: branch \"%s\" requires a mode\n",
8624 + optarg);
8625 + goto out;
8626 + }
8627 + *modename++ = '\0';
8628 + err = parse_branch_mode(modename, &perms);
8629 + if (err) {
8630 + printk(KERN_ERR "unionfs: invalid mode \"%s\" for \"%s\"\n",
8631 + modename, optarg);
8632 + goto out;
8633 + }
8634 +
8635 + /*
8636 + * Find matching branch index. For now, this assumes that nothing
8637 + * has been mounted on top of this Unionfs stack. Once we have /odf
8638 + * and cache-coherency resolved, we'll address the branch-path
8639 + * uniqueness.
8640 + */
8641 + err = path_lookup(optarg, LOOKUP_FOLLOW, &nd);
8642 + if (err) {
8643 + printk(KERN_ERR "unionfs: error accessing "
8644 + "lower directory \"%s\" (error %d)\n",
8645 + optarg, err);
8646 + goto out;
8647 + }
8648 + for (idx = 0; idx < cur_branches; idx++)
8649 + if (nd.path.mnt == new_lower_paths[idx].mnt &&
8650 + nd.path.dentry == new_lower_paths[idx].dentry)
8651 + break;
8652 + path_put(&nd.path); /* no longer needed */
8653 + if (idx == cur_branches) {
8654 + err = -ENOENT; /* err may have been reset above */
8655 + printk(KERN_ERR "unionfs: branch \"%s\" "
8656 + "not found\n", optarg);
8657 + goto out;
8658 + }
8659 + /* check/change mode for existing branch */
8660 + /* we don't warn if perms==branchperms */
8661 + new_data[idx].branchperms = perms;
8662 + err = 0;
8663 +out:
8664 + return err;
8665 +}
8666 +
8667 +/* handle branch deletion during remount */
8668 +static noinline_for_stack int do_remount_del_option(
8669 + char *optarg, int cur_branches,
8670 + struct unionfs_data *new_data,
8671 + struct path *new_lower_paths)
8672 +{
8673 + int err = -EINVAL;
8674 + int idx;
8675 + struct nameidata nd;
8676 +
8677 + /* optarg contains the branch name to delete */
8678 +
8679 + /*
8680 + * Find matching branch index. For now, this assumes that nothing
8681 + * has been mounted on top of this Unionfs stack. Once we have /odf
8682 + * and cache-coherency resolved, we'll address the branch-path
8683 + * uniqueness.
8684 + */
8685 + err = path_lookup(optarg, LOOKUP_FOLLOW, &nd);
8686 + if (err) {
8687 + printk(KERN_ERR "unionfs: error accessing "
8688 + "lower directory \"%s\" (error %d)\n",
8689 + optarg, err);
8690 + goto out;
8691 + }
8692 + for (idx = 0; idx < cur_branches; idx++)
8693 + if (nd.path.mnt == new_lower_paths[idx].mnt &&
8694 + nd.path.dentry == new_lower_paths[idx].dentry)
8695 + break;
8696 + path_put(&nd.path); /* no longer needed */
8697 + if (idx == cur_branches) {
8698 + printk(KERN_ERR "unionfs: branch \"%s\" "
8699 + "not found\n", optarg);
8700 + err = -ENOENT;
8701 + goto out;
8702 + }
8703 + /* check if there are any open files on the branch to be deleted */
8704 + if (atomic_read(&new_data[idx].open_files) > 0) {
8705 + err = -EBUSY;
8706 + goto out;
8707 + }
8708 +
8709 + /*
8710 + * Now we have to delete the branch. First, release any handles it
8711 + * has. Then, move the remaining array indexes past "idx" in
8712 + * new_data and new_lower_paths one to the left. Finally, adjust
8713 + * cur_branches.
8714 + */
8715 + path_put(&new_lower_paths[idx]);
8716 +
8717 + if (idx < cur_branches - 1) {
8718 + /* if idx==cur_branches-1, we delete last branch: easy */
8719 + memmove(&new_data[idx], &new_data[idx+1],
8720 + (cur_branches - 1 - idx) *
8721 + sizeof(struct unionfs_data));
8722 + memmove(&new_lower_paths[idx], &new_lower_paths[idx+1],
8723 + (cur_branches - 1 - idx) * sizeof(struct path));
8724 + }
8725 +
8726 + err = 0;
8727 +out:
8728 + return err;
8729 +}
8730 +
8731 +/* handle branch insertion during remount */
8732 +static noinline_for_stack int do_remount_add_option(
8733 + char *optarg, int cur_branches,
8734 + struct unionfs_data *new_data,
8735 + struct path *new_lower_paths,
8736 + int *high_branch_id)
8737 +{
8738 + int err = -EINVAL;
8739 + int perms;
8740 + int idx = 0; /* default: insert at beginning */
8741 + char *new_branch , *modename = NULL;
8742 + struct nameidata nd;
8743 +
8744 + /*
8745 + * optarg can be of several forms:
8746 + *
8747 + * /bar:/foo insert /foo before /bar
8748 + * /bar:/foo=ro insert /foo in ro mode before /bar
8749 + * /foo insert /foo in the beginning (prepend)
8750 + * :/foo insert /foo at the end (append)
8751 + */
8752 + if (*optarg == ':') { /* append? */
8753 + new_branch = optarg + 1; /* skip ':' */
8754 + idx = cur_branches;
8755 + goto found_insertion_point;
8756 + }
8757 + new_branch = strchr(optarg, ':');
8758 + if (!new_branch) { /* prepend? */
8759 + new_branch = optarg;
8760 + goto found_insertion_point;
8761 + }
8762 + *new_branch++ = '\0'; /* holds path+mode of new branch */
8763 +
8764 + /*
8765 + * Find matching branch index. For now, this assumes that nothing
8766 + * has been mounted on top of this Unionfs stack. Once we have /odf
8767 + * and cache-coherency resolved, we'll address the branch-path
8768 + * uniqueness.
8769 + */
8770 + err = path_lookup(optarg, LOOKUP_FOLLOW, &nd);
8771 + if (err) {
8772 + printk(KERN_ERR "unionfs: error accessing "
8773 + "lower directory \"%s\" (error %d)\n",
8774 + optarg, err);
8775 + goto out;
8776 + }
8777 + for (idx = 0; idx < cur_branches; idx++)
8778 + if (nd.path.mnt == new_lower_paths[idx].mnt &&
8779 + nd.path.dentry == new_lower_paths[idx].dentry)
8780 + break;
8781 + path_put(&nd.path); /* no longer needed */
8782 + if (idx == cur_branches) {
8783 + printk(KERN_ERR "unionfs: branch \"%s\" "
8784 + "not found\n", optarg);
8785 + err = -ENOENT;
8786 + goto out;
8787 + }
8788 +
8789 + /*
8790 + * At this point idx will hold the index where the new branch should
8791 + * be inserted before.
8792 + */
8793 +found_insertion_point:
8794 + /* find the mode for the new branch */
8795 + if (new_branch)
8796 + modename = strchr(new_branch, '=');
8797 + if (modename)
8798 + *modename++ = '\0';
8799 + if (!new_branch || !*new_branch) {
8800 + printk(KERN_ERR "unionfs: null new branch\n");
8801 + err = -EINVAL;
8802 + goto out;
8803 + }
8804 + err = parse_branch_mode(modename, &perms);
8805 + if (err) {
8806 + printk(KERN_ERR "unionfs: invalid mode \"%s\" for "
8807 + "branch \"%s\"\n", modename, new_branch);
8808 + goto out;
8809 + }
8810 + err = path_lookup(new_branch, LOOKUP_FOLLOW, &nd);
8811 + if (err) {
8812 + printk(KERN_ERR "unionfs: error accessing "
8813 + "lower directory \"%s\" (error %d)\n",
8814 + new_branch, err);
8815 + goto out;
8816 + }
8817 + /*
8818 + * It's probably safe to check_mode the new branch to insert. Note:
8819 + * we don't allow inserting branches which are unionfs's by
8820 + * themselves (check_branch returns EINVAL in that case). This is
8821 + * because this code base doesn't support stacking unionfs: the ODF
8822 + * code base supports that correctly.
8823 + */
8824 + err = check_branch(&nd);
8825 + if (err) {
8826 + printk(KERN_ERR "unionfs: lower directory "
8827 + "\"%s\" is not a valid branch\n", optarg);
8828 + path_put(&nd.path);
8829 + goto out;
8830 + }
8831 +
8832 + /*
8833 + * Now we have to insert the new branch. But first, move the bits
8834 + * to make space for the new branch, if needed. Finally, adjust
8835 + * cur_branches.
8836 + * We don't release nd here; it's kept until umount/remount.
8837 + */
8838 + if (idx < cur_branches) {
8839 + /* if idx==cur_branches, we append: easy */
8840 + memmove(&new_data[idx+1], &new_data[idx],
8841 + (cur_branches - idx) * sizeof(struct unionfs_data));
8842 + memmove(&new_lower_paths[idx+1], &new_lower_paths[idx],
8843 + (cur_branches - idx) * sizeof(struct path));
8844 + }
8845 + new_lower_paths[idx].dentry = nd.path.dentry;
8846 + new_lower_paths[idx].mnt = nd.path.mnt;
8847 +
8848 + new_data[idx].sb = nd.path.dentry->d_sb;
8849 + atomic_set(&new_data[idx].open_files, 0);
8850 + new_data[idx].branchperms = perms;
8851 + new_data[idx].branch_id = ++*high_branch_id; /* assign new branch ID */
8852 +
8853 + err = 0;
8854 +out:
8855 + return err;
8856 +}
8857 +
8858 +
8859 +/*
8860 + * Support branch management options on remount.
8861 + *
8862 + * See Documentation/filesystems/unionfs/ for details.
8863 + *
8864 + * @flags: numeric mount options
8865 + * @options: mount options string
8866 + *
8867 + * This function can rearrange a mounted union dynamically, adding and
8868 + * removing branches, including changing branch modes. Clearly this has to
8869 + * be done safely and atomically. Luckily, the VFS already calls this
8870 + * function with lock_super(sb) and lock_kernel() held, preventing
8871 + * concurrent mixing of new mounts, remounts, and unmounts. Moreover,
8872 + * do_remount_sb(), our caller function, already called shrink_dcache_sb(sb)
8873 + * to purge dentries/inodes from our superblock, and also called
8874 + * fsync_super(sb) to purge any dirty pages. So we're good.
8875 + *
8876 + * XXX: however, our remount code may also need to invalidate mapped pages
8877 + * so as to force them to be re-gotten from the (newly reconfigured) lower
8878 + * branches. This has to wait for proper mmap and cache coherency support
8879 + * in the VFS.
8880 + *
8881 + */
8882 +static int unionfs_remount_fs(struct super_block *sb, int *flags,
8883 + char *options)
8884 +{
8885 + int err = 0;
8886 + int i;
8887 + char *optionstmp, *tmp_to_free; /* kstrdup'ed of "options" */
8888 + char *optname;
8889 + int cur_branches = 0; /* no. of current branches */
8890 + int new_branches = 0; /* no. of branches actually left in the end */
8891 + int add_branches; /* est. no. of branches to add */
8892 + int del_branches; /* est. no. of branches to del */
8893 + int max_branches; /* max possible no. of branches */
8894 + struct unionfs_data *new_data = NULL, *tmp_data = NULL;
8895 + struct path *new_lower_paths = NULL, *tmp_lower_paths = NULL;
8896 + struct inode **new_lower_inodes = NULL;
8897 + int new_high_branch_id; /* new high branch ID */
8898 + int size; /* memory allocation size, temp var */
8899 + int old_ibstart, old_ibend;
8900 +
8901 + unionfs_write_lock(sb);
8902 +
8903 + /*
8904 + * The VFS will take care of "ro" and "rw" flags, and we can safely
8905 + * ignore MS_SILENT, but anything else left over is an error. So we
8906 + * need to check if any other flags may have been passed (none are
8907 + * allowed/supported as of now).
8908 + */
8909 + if ((*flags & ~(MS_RDONLY | MS_SILENT)) != 0) {
8910 + printk(KERN_ERR
8911 + "unionfs: remount flags 0x%x unsupported\n", *flags);
8912 + err = -EINVAL;
8913 + goto out_error;
8914 + }
8915 +
8916 + /*
8917 + * If 'options' is NULL, it's probably because the user just changed
8918 + * the union to a "ro" or "rw" and the VFS took care of it. So
8919 + * nothing to do and we're done.
8920 + */
8921 + if (!options || options[0] == '\0')
8922 + goto out_error;
8923 +
8924 + /*
8925 + * Find out how many branches we will have in the end, counting
8926 + * "add" and "del" commands. Copy the "options" string because
8927 + * strsep modifies the string and we need it later.
8928 + */
8929 + tmp_to_free = kstrdup(options, GFP_KERNEL);
8930 + optionstmp = tmp_to_free;
8931 + if (unlikely(!optionstmp)) {
8932 + err = -ENOMEM;
8933 + goto out_free;
8934 + }
8935 + cur_branches = sbmax(sb); /* current no. branches */
8936 + new_branches = sbmax(sb);
8937 + del_branches = 0;
8938 + add_branches = 0;
8939 + new_high_branch_id = sbhbid(sb); /* save current high_branch_id */
8940 + while ((optname = strsep(&optionstmp, ",")) != NULL) {
8941 + char *optarg;
8942 +
8943 + if (!optname || !*optname)
8944 + continue;
8945 +
8946 + optarg = strchr(optname, '=');
8947 + if (optarg)
8948 + *optarg++ = '\0';
8949 +
8950 + if (!strcmp("add", optname))
8951 + add_branches++;
8952 + else if (!strcmp("del", optname))
8953 + del_branches++;
8954 + }
8955 + kfree(tmp_to_free);
8956 + /* after all changes, will we have at least one branch left? */
8957 + if ((new_branches + add_branches - del_branches) < 1) {
8958 + printk(KERN_ERR
8959 + "unionfs: no branches left after remount\n");
8960 + err = -EINVAL;
8961 + goto out_free;
8962 + }
8963 +
8964 + /*
8965 + * Since we haven't actually parsed all the add/del options, nor
8966 + * have we checked them for errors, we don't know for sure how many
8967 + * branches we will have after all changes have taken place. In
8968 + * fact, the total number of branches left could be less than what
8969 + * we have now. So we need to allocate space for a temporary
8970 + * placeholder that is at least as large as the maximum number of
8971 + * branches we *could* have, which is the current number plus all
8972 + * the additions. Once we're done with these temp placeholders, we
8973 + * may have to re-allocate the final size, copy over from the temp,
8974 + * and then free the temps (done near the end of this function).
8975 + */
8976 + max_branches = cur_branches + add_branches;
8977 + /* allocate space for new pointers to lower dentry */
8978 + tmp_data = kcalloc(max_branches,
8979 + sizeof(struct unionfs_data), GFP_KERNEL);
8980 + if (unlikely(!tmp_data)) {
8981 + err = -ENOMEM;
8982 + goto out_free;
8983 + }
8984 + /* allocate space for new pointers to lower paths */
8985 + tmp_lower_paths = kcalloc(max_branches,
8986 + sizeof(struct path), GFP_KERNEL);
8987 + if (unlikely(!tmp_lower_paths)) {
8988 + err = -ENOMEM;
8989 + goto out_free;
8990 + }
8991 + /* copy current info into new placeholders, incrementing refcnts */
8992 + memcpy(tmp_data, UNIONFS_SB(sb)->data,
8993 + cur_branches * sizeof(struct unionfs_data));
8994 + memcpy(tmp_lower_paths, UNIONFS_D(sb->s_root)->lower_paths,
8995 + cur_branches * sizeof(struct path));
8996 + for (i = 0; i < cur_branches; i++)
8997 + path_get(&tmp_lower_paths[i]); /* drop refs at end of fxn */
8998 +
8999 + /*******************************************************************
9000 + * For each branch command, do path_lookup on the requested branch,
9001 + * and apply the change to a temp branch list. To handle errors, we
9002 + * already dup'ed the old arrays (above), and increased the refcnts
9003 + * on various f/s objects. So now we can do all the path_lookups
9004 + * and branch-management commands on the new arrays. If it fail mid
9005 + * way, we free the tmp arrays and *put all objects. If we succeed,
9006 + * then we free old arrays and *put its objects, and then replace
9007 + * the arrays with the new tmp list (we may have to re-allocate the
9008 + * memory because the temp lists could have been larger than what we
9009 + * actually needed).
9010 + *******************************************************************/
9011 +
9012 + while ((optname = strsep(&options, ",")) != NULL) {
9013 + char *optarg;
9014 +
9015 + if (!optname || !*optname)
9016 + continue;
9017 + /*
9018 + * At this stage optname holds a comma-delimited option, but
9019 + * without the commas. Next, we need to break the string on
9020 + * the '=' symbol to separate CMD=ARG, where ARG itself can
9021 + * be KEY=VAL. For example, in mode=/foo=rw, CMD is "mode",
9022 + * KEY is "/foo", and VAL is "rw".
9023 + */
9024 + optarg = strchr(optname, '=');
9025 + if (optarg)
9026 + *optarg++ = '\0';
9027 + /* incgen remount option (instead of old ioctl) */
9028 + if (!strcmp("incgen", optname)) {
9029 + err = 0;
9030 + goto out_no_change;
9031 + }
9032 +
9033 + /*
9034 + * All of our options take an argument now. (Insert ones
9035 + * that don't above this check.) So at this stage optname
9036 + * contains the CMD part and optarg contains the ARG part.
9037 + */
9038 + if (!optarg || !*optarg) {
9039 + printk(KERN_ERR "unionfs: all remount options require "
9040 + "an argument (%s)\n", optname);
9041 + err = -EINVAL;
9042 + goto out_release;
9043 + }
9044 +
9045 + if (!strcmp("add", optname)) {
9046 + err = do_remount_add_option(optarg, new_branches,
9047 + tmp_data,
9048 + tmp_lower_paths,
9049 + &new_high_branch_id);
9050 + if (err)
9051 + goto out_release;
9052 + new_branches++;
9053 + if (new_branches > UNIONFS_MAX_BRANCHES) {
9054 + printk(KERN_ERR "unionfs: command exceeds "
9055 + "%d branches\n", UNIONFS_MAX_BRANCHES);
9056 + err = -E2BIG;
9057 + goto out_release;
9058 + }
9059 + continue;
9060 + }
9061 + if (!strcmp("del", optname)) {
9062 + err = do_remount_del_option(optarg, new_branches,
9063 + tmp_data,
9064 + tmp_lower_paths);
9065 + if (err)
9066 + goto out_release;
9067 + new_branches--;
9068 + continue;
9069 + }
9070 + if (!strcmp("mode", optname)) {
9071 + err = do_remount_mode_option(optarg, new_branches,
9072 + tmp_data,
9073 + tmp_lower_paths);
9074 + if (err)
9075 + goto out_release;
9076 + continue;
9077 + }
9078 +
9079 + /*
9080 + * When you use "mount -o remount,ro", mount(8) will
9081 + * reportedly pass the original dirs= string from
9082 + * /proc/mounts. So for now, we have to ignore dirs= and
9083 + * not consider it an error, unless we want to allow users
9084 + * to pass dirs= in remount. Note that to allow the VFS to
9085 + * actually process the ro/rw remount options, we have to
9086 + * return 0 from this function.
9087 + */
9088 + if (!strcmp("dirs", optname)) {
9089 + printk(KERN_WARNING
9090 + "unionfs: remount ignoring option \"%s\"\n",
9091 + optname);
9092 + continue;
9093 + }
9094 +
9095 + err = -EINVAL;
9096 + printk(KERN_ERR
9097 + "unionfs: unrecognized option \"%s\"\n", optname);
9098 + goto out_release;
9099 + }
9100 +
9101 +out_no_change:
9102 +
9103 + /******************************************************************
9104 + * WE'RE ALMOST DONE: check if leftmost branch might be read-only,
9105 + * see if we need to allocate a small-sized new vector, copy the
9106 + * vectors to their correct place, release the refcnt of the older
9107 + * ones, and return. Also handle invalidating any pages that will
9108 + * have to be re-read.
9109 + *******************************************************************/
9110 +
9111 + if (!(tmp_data[0].branchperms & MAY_WRITE)) {
9112 + printk(KERN_ERR "unionfs: leftmost branch cannot be read-only "
9113 + "(use \"remount,ro\" to create a read-only union)\n");
9114 + err = -EINVAL;
9115 + goto out_release;
9116 + }
9117 +
9118 + /* (re)allocate space for new pointers to lower dentry */
9119 + size = new_branches * sizeof(struct unionfs_data);
9120 + new_data = krealloc(tmp_data, size, GFP_KERNEL);
9121 + if (unlikely(!new_data)) {
9122 + err = -ENOMEM;
9123 + goto out_release;
9124 + }
9125 +
9126 + /* allocate space for new pointers to lower paths */
9127 + size = new_branches * sizeof(struct path);
9128 + new_lower_paths = krealloc(tmp_lower_paths, size, GFP_KERNEL);
9129 + if (unlikely(!new_lower_paths)) {
9130 + err = -ENOMEM;
9131 + goto out_release;
9132 + }
9133 +
9134 + /* allocate space for new pointers to lower inodes */
9135 + new_lower_inodes = kcalloc(new_branches,
9136 + sizeof(struct inode *), GFP_KERNEL);
9137 + if (unlikely(!new_lower_inodes)) {
9138 + err = -ENOMEM;
9139 + goto out_release;
9140 + }
9141 +
9142 + /*
9143 + * OK, just before we actually put the new set of branches in place,
9144 + * we need to ensure that our own f/s has no dirty objects left.
9145 + * Luckily, do_remount_sb() already calls shrink_dcache_sb(sb) and
9146 + * fsync_super(sb), taking care of dentries, inodes, and dirty
9147 + * pages. So all that's left is for us to invalidate any leftover
9148 + * (non-dirty) pages to ensure that they will be re-read from the
9149 + * new lower branches (and to support mmap).
9150 + */
9151 +
9152 + /*
9153 + * Once we finish the remounting successfully, our superblock
9154 + * generation number will have increased. This will be detected by
9155 + * our dentry-revalidation code upon subsequent f/s operations
9156 + * through unionfs. The revalidation code will rebuild the union of
9157 + * lower inodes for a given unionfs inode and invalidate any pages
9158 + * of such "stale" inodes (by calling our purge_inode_data
9159 + * function). This revalidation will happen lazily and
9160 + * incrementally, as users perform operations on cached inodes. We
9161 + * would like to encourage this revalidation to happen sooner if
9162 + * possible, so we like to try to invalidate as many other pages in
9163 + * our superblock as we can. We used to call drop_pagecache_sb() or
9164 + * a variant thereof, but either method was racy (drop_caches alone
9165 + * is known to be racy). So now we let the revalidation happen on a
9166 + * per file basis in ->d_revalidate.
9167 + */
9168 +
9169 + /* grab new lower super references; release old ones */
9170 + for (i = 0; i < new_branches; i++)
9171 + atomic_inc(&new_data[i].sb->s_active);
9172 + for (i = 0; i < sbmax(sb); i++)
9173 + atomic_dec(&UNIONFS_SB(sb)->data[i].sb->s_active);
9174 +
9175 + /* copy new vectors into their correct place */
9176 + tmp_data = UNIONFS_SB(sb)->data;
9177 + UNIONFS_SB(sb)->data = new_data;
9178 + new_data = NULL; /* so don't free good pointers below */
9179 + tmp_lower_paths = UNIONFS_D(sb->s_root)->lower_paths;
9180 + UNIONFS_D(sb->s_root)->lower_paths = new_lower_paths;
9181 + new_lower_paths = NULL; /* so don't free good pointers below */
9182 +
9183 + /* update our unionfs_sb_info and root dentry index of last branch */
9184 + i = sbmax(sb); /* save no. of branches to release at end */
9185 + sbend(sb) = new_branches - 1;
9186 + dbend(sb->s_root) = new_branches - 1;
9187 + old_ibstart = ibstart(sb->s_root->d_inode);
9188 + old_ibend = ibend(sb->s_root->d_inode);
9189 + ibend(sb->s_root->d_inode) = new_branches - 1;
9190 + UNIONFS_D(sb->s_root)->bcount = new_branches;
9191 + new_branches = i; /* no. of branches to release below */
9192 +
9193 + /*
9194 + * Update lower inodes: 3 steps
9195 + * 1. grab ref on all new lower inodes
9196 + */
9197 + for (i = dbstart(sb->s_root); i <= dbend(sb->s_root); i++) {
9198 + struct dentry *lower_dentry =
9199 + unionfs_lower_dentry_idx(sb->s_root, i);
9200 + igrab(lower_dentry->d_inode);
9201 + new_lower_inodes[i] = lower_dentry->d_inode;
9202 + }
9203 + /* 2. release reference on all older lower inodes */
9204 + iput_lowers(sb->s_root->d_inode, old_ibstart, old_ibend, true);
9205 + /* 3. update root dentry's inode to new lower_inodes array */
9206 + UNIONFS_I(sb->s_root->d_inode)->lower_inodes = new_lower_inodes;
9207 + new_lower_inodes = NULL;
9208 +
9209 + /* maxbytes may have changed */
9210 + sb->s_maxbytes = unionfs_lower_super_idx(sb, 0)->s_maxbytes;
9211 + /* update high branch ID */
9212 + sbhbid(sb) = new_high_branch_id;
9213 +
9214 + /* update our sb->generation for revalidating objects */
9215 + i = atomic_inc_return(&UNIONFS_SB(sb)->generation);
9216 + atomic_set(&UNIONFS_D(sb->s_root)->generation, i);
9217 + atomic_set(&UNIONFS_I(sb->s_root->d_inode)->generation, i);
9218 + if (!(*flags & MS_SILENT))
9219 + pr_info("unionfs: %s: new generation number %d\n",
9220 + UNIONFS_SB(sb)->dev_name, i);
9221 + /* finally, update the root dentry's times */
9222 + unionfs_copy_attr_times(sb->s_root->d_inode);
9223 + err = 0; /* reset to success */
9224 +
9225 + /*
9226 + * The code above falls through to the next label, and releases the
9227 + * refcnts of the older ones (stored in tmp_*): if we fell through
9228 + * here, it means success. However, if we jump directly to this
9229 + * label from any error above, then an error occurred after we
9230 + * grabbed various refcnts, and so we have to release the
9231 + * temporarily constructed structures.
9232 + */
9233 +out_release:
9234 + /* no need to cleanup/release anything in tmp_data */
9235 + if (tmp_lower_paths)
9236 + for (i = 0; i < new_branches; i++)
9237 + path_put(&tmp_lower_paths[i]);
9238 +out_free:
9239 + kfree(tmp_lower_paths);
9240 + kfree(tmp_data);
9241 + kfree(new_lower_paths);
9242 + kfree(new_data);
9243 + kfree(new_lower_inodes);
9244 +out_error:
9245 + unionfs_check_dentry(sb->s_root);
9246 + unionfs_write_unlock(sb);
9247 + return err;
9248 +}
9249 +
9250 +/*
9251 + * Called by iput() when the inode reference count reached zero
9252 + * and the inode is not hashed anywhere. Used to clear anything
9253 + * that needs to be, before the inode is completely destroyed and put
9254 + * on the inode free list.
9255 + *
9256 + * No need to lock sb info's rwsem.
9257 + */
9258 +static void unionfs_clear_inode(struct inode *inode)
9259 +{
9260 + int bindex, bstart, bend;
9261 + struct inode *lower_inode;
9262 + struct list_head *pos, *n;
9263 + struct unionfs_dir_state *rdstate;
9264 +
9265 + list_for_each_safe(pos, n, &UNIONFS_I(inode)->readdircache) {
9266 + rdstate = list_entry(pos, struct unionfs_dir_state, cache);
9267 + list_del(&rdstate->cache);
9268 + free_rdstate(rdstate);
9269 + }
9270 +
9271 + /*
9272 + * Decrement a reference to a lower_inode, which was incremented
9273 + * by our read_inode when it was created initially.
9274 + */
9275 + bstart = ibstart(inode);
9276 + bend = ibend(inode);
9277 + if (bstart >= 0) {
9278 + for (bindex = bstart; bindex <= bend; bindex++) {
9279 + lower_inode = unionfs_lower_inode_idx(inode, bindex);
9280 + if (!lower_inode)
9281 + continue;
9282 + unionfs_set_lower_inode_idx(inode, bindex, NULL);
9283 + /* see Documentation/filesystems/unionfs/issues.txt */
9284 + lockdep_off();
9285 + iput(lower_inode);
9286 + lockdep_on();
9287 + }
9288 + }
9289 +
9290 + kfree(UNIONFS_I(inode)->lower_inodes);
9291 + UNIONFS_I(inode)->lower_inodes = NULL;
9292 +}
9293 +
9294 +static struct inode *unionfs_alloc_inode(struct super_block *sb)
9295 +{
9296 + struct unionfs_inode_info *i;
9297 +
9298 + i = kmem_cache_alloc(unionfs_inode_cachep, GFP_KERNEL);
9299 + if (unlikely(!i))
9300 + return NULL;
9301 +
9302 + /* memset everything up to the inode to 0 */
9303 + memset(i, 0, offsetof(struct unionfs_inode_info, vfs_inode));
9304 +
9305 + i->vfs_inode.i_version = 1;
9306 + return &i->vfs_inode;
9307 +}
9308 +
9309 +static void unionfs_destroy_inode(struct inode *inode)
9310 +{
9311 + kmem_cache_free(unionfs_inode_cachep, UNIONFS_I(inode));
9312 +}
9313 +
9314 +/* unionfs inode cache constructor */
9315 +static void init_once(void *obj)
9316 +{
9317 + struct unionfs_inode_info *i = obj;
9318 +
9319 + inode_init_once(&i->vfs_inode);
9320 +}
9321 +
9322 +int unionfs_init_inode_cache(void)
9323 +{
9324 + int err = 0;
9325 +
9326 + unionfs_inode_cachep =
9327 + kmem_cache_create("unionfs_inode_cache",
9328 + sizeof(struct unionfs_inode_info), 0,
9329 + SLAB_RECLAIM_ACCOUNT, init_once);
9330 + if (unlikely(!unionfs_inode_cachep))
9331 + err = -ENOMEM;
9332 + return err;
9333 +}
9334 +
9335 +/* unionfs inode cache destructor */
9336 +void unionfs_destroy_inode_cache(void)
9337 +{
9338 + if (unionfs_inode_cachep)
9339 + kmem_cache_destroy(unionfs_inode_cachep);
9340 +}
9341 +
9342 +/*
9343 + * Called when we have a dirty inode, right here we only throw out
9344 + * parts of our readdir list that are too old.
9345 + *
9346 + * No need to grab sb info's rwsem.
9347 + */
9348 +static int unionfs_write_inode(struct inode *inode, int sync)
9349 +{
9350 + struct list_head *pos, *n;
9351 + struct unionfs_dir_state *rdstate;
9352 +
9353 + spin_lock(&UNIONFS_I(inode)->rdlock);
9354 + list_for_each_safe(pos, n, &UNIONFS_I(inode)->readdircache) {
9355 + rdstate = list_entry(pos, struct unionfs_dir_state, cache);
9356 + /* We keep this list in LRU order. */
9357 + if ((rdstate->access + RDCACHE_JIFFIES) > jiffies)
9358 + break;
9359 + UNIONFS_I(inode)->rdcount--;
9360 + list_del(&rdstate->cache);
9361 + free_rdstate(rdstate);
9362 + }
9363 + spin_unlock(&UNIONFS_I(inode)->rdlock);
9364 +
9365 + return 0;
9366 +}
9367 +
9368 +/*
9369 + * Used only in nfs, to kill any pending RPC tasks, so that subsequent
9370 + * code can actually succeed and won't leave tasks that need handling.
9371 + */
9372 +static void unionfs_umount_begin(struct super_block *sb)
9373 +{
9374 + struct super_block *lower_sb;
9375 + int bindex, bstart, bend;
9376 +
9377 + unionfs_read_lock(sb, UNIONFS_SMUTEX_CHILD);
9378 +
9379 + bstart = sbstart(sb);
9380 + bend = sbend(sb);
9381 + for (bindex = bstart; bindex <= bend; bindex++) {
9382 + lower_sb = unionfs_lower_super_idx(sb, bindex);
9383 +
9384 + if (lower_sb && lower_sb->s_op &&
9385 + lower_sb->s_op->umount_begin)
9386 + lower_sb->s_op->umount_begin(lower_sb);
9387 + }
9388 +
9389 + unionfs_read_unlock(sb);
9390 +}
9391 +
9392 +static int unionfs_show_options(struct seq_file *m, struct vfsmount *mnt)
9393 +{
9394 + struct super_block *sb = mnt->mnt_sb;
9395 + int ret = 0;
9396 + char *tmp_page;
9397 + char *path;
9398 + int bindex, bstart, bend;
9399 + int perms;
9400 +
9401 + unionfs_read_lock(sb, UNIONFS_SMUTEX_CHILD);
9402 +
9403 + unionfs_lock_dentry(sb->s_root, UNIONFS_DMUTEX_CHILD);
9404 +
9405 + tmp_page = (char *) __get_free_page(GFP_KERNEL);
9406 + if (unlikely(!tmp_page)) {
9407 + ret = -ENOMEM;
9408 + goto out;
9409 + }
9410 +
9411 + bstart = sbstart(sb);
9412 + bend = sbend(sb);
9413 +
9414 + seq_printf(m, ",dirs=");
9415 + for (bindex = bstart; bindex <= bend; bindex++) {
9416 + struct path p;
9417 + p.dentry = unionfs_lower_dentry_idx(sb->s_root, bindex);
9418 + p.mnt = unionfs_lower_mnt_idx(sb->s_root, bindex);
9419 + path = d_path(&p, tmp_page, PAGE_SIZE);
9420 + if (IS_ERR(path)) {
9421 + ret = PTR_ERR(path);
9422 + goto out;
9423 + }
9424 +
9425 + perms = branchperms(sb, bindex);
9426 +
9427 + seq_printf(m, "%s=%s", path,
9428 + perms & MAY_WRITE ? "rw" : "ro");
9429 + if (bindex != bend)
9430 + seq_printf(m, ":");
9431 + }
9432 +
9433 +out:
9434 + free_page((unsigned long) tmp_page);
9435 +
9436 + unionfs_unlock_dentry(sb->s_root);
9437 +
9438 + unionfs_read_unlock(sb);
9439 +
9440 + return ret;
9441 +}
9442 +
9443 +struct super_operations unionfs_sops = {
9444 + .delete_inode = unionfs_delete_inode,
9445 + .put_super = unionfs_put_super,
9446 + .statfs = unionfs_statfs,
9447 + .remount_fs = unionfs_remount_fs,
9448 + .clear_inode = unionfs_clear_inode,
9449 + .umount_begin = unionfs_umount_begin,
9450 + .show_options = unionfs_show_options,
9451 + .write_inode = unionfs_write_inode,
9452 + .alloc_inode = unionfs_alloc_inode,
9453 + .destroy_inode = unionfs_destroy_inode,
9454 +};
9455 diff -Naur linux-2.6.30/fs/unionfs/union.h linux-2.6.30-magellan/fs/unionfs/union.h
9456 --- linux-2.6.30/fs/unionfs/union.h 1970-01-01 01:00:00.000000000 +0100
9457 +++ linux-2.6.30-magellan/fs/unionfs/union.h 2009-06-11 18:39:19.000000000 +0200
9458 @@ -0,0 +1,650 @@
9459 +/*
9460 + * Copyright (c) 2003-2009 Erez Zadok
9461 + * Copyright (c) 2003-2006 Charles P. Wright
9462 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
9463 + * Copyright (c) 2005 Arun M. Krishnakumar
9464 + * Copyright (c) 2004-2006 David P. Quigley
9465 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
9466 + * Copyright (c) 2003 Puja Gupta
9467 + * Copyright (c) 2003 Harikesavan Krishnan
9468 + * Copyright (c) 2003-2009 Stony Brook University
9469 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
9470 + *
9471 + * This program is free software; you can redistribute it and/or modify
9472 + * it under the terms of the GNU General Public License version 2 as
9473 + * published by the Free Software Foundation.
9474 + */
9475 +
9476 +#ifndef _UNION_H_
9477 +#define _UNION_H_
9478 +
9479 +#include <linux/dcache.h>
9480 +#include <linux/file.h>
9481 +#include <linux/list.h>
9482 +#include <linux/fs.h>
9483 +#include <linux/mm.h>
9484 +#include <linux/module.h>
9485 +#include <linux/mount.h>
9486 +#include <linux/namei.h>
9487 +#include <linux/page-flags.h>
9488 +#include <linux/pagemap.h>
9489 +#include <linux/poll.h>
9490 +#include <linux/security.h>
9491 +#include <linux/seq_file.h>
9492 +#include <linux/slab.h>
9493 +#include <linux/spinlock.h>
9494 +#include <linux/smp_lock.h>
9495 +#include <linux/statfs.h>
9496 +#include <linux/string.h>
9497 +#include <linux/vmalloc.h>
9498 +#include <linux/writeback.h>
9499 +#include <linux/buffer_head.h>
9500 +#include <linux/xattr.h>
9501 +#include <linux/fs_stack.h>
9502 +#include <linux/magic.h>
9503 +#include <linux/log2.h>
9504 +#include <linux/poison.h>
9505 +#include <linux/mman.h>
9506 +#include <linux/backing-dev.h>
9507 +#include <linux/splice.h>
9508 +
9509 +#include <asm/system.h>
9510 +
9511 +#include <linux/union_fs.h>
9512 +
9513 +/* the file system name */
9514 +#define UNIONFS_NAME "unionfs"
9515 +
9516 +/* unionfs root inode number */
9517 +#define UNIONFS_ROOT_INO 1
9518 +
9519 +/* number of times we try to get a unique temporary file name */
9520 +#define GET_TMPNAM_MAX_RETRY 5
9521 +
9522 +/* maximum number of branches we support, to avoid memory blowup */
9523 +#define UNIONFS_MAX_BRANCHES 128
9524 +
9525 +/* minimum time (seconds) required for time-based cache-coherency */
9526 +#define UNIONFS_MIN_CC_TIME 3
9527 +
9528 +/* Operations vectors defined in specific files. */
9529 +extern struct file_operations unionfs_main_fops;
9530 +extern struct file_operations unionfs_dir_fops;
9531 +extern struct inode_operations unionfs_main_iops;
9532 +extern struct inode_operations unionfs_dir_iops;
9533 +extern struct inode_operations unionfs_symlink_iops;
9534 +extern struct super_operations unionfs_sops;
9535 +extern struct dentry_operations unionfs_dops;
9536 +extern struct address_space_operations unionfs_aops, unionfs_dummy_aops;
9537 +extern struct vm_operations_struct unionfs_vm_ops;
9538 +
9539 +/* How long should an entry be allowed to persist */
9540 +#define RDCACHE_JIFFIES (5*HZ)
9541 +
9542 +/* compatibility with Real-Time patches */
9543 +#ifdef CONFIG_PREEMPT_RT
9544 +# define unionfs_rw_semaphore compat_rw_semaphore
9545 +#else /* not CONFIG_PREEMPT_RT */
9546 +# define unionfs_rw_semaphore rw_semaphore
9547 +#endif /* not CONFIG_PREEMPT_RT */
9548 +
9549 +/* file private data. */
9550 +struct unionfs_file_info {
9551 + int bstart;
9552 + int bend;
9553 + atomic_t generation;
9554 +
9555 + struct unionfs_dir_state *rdstate;
9556 + struct file **lower_files;
9557 + int *saved_branch_ids; /* IDs of branches when file was opened */
9558 + struct vm_operations_struct *lower_vm_ops;
9559 + bool wrote_to_file; /* for delayed copyup */
9560 +};
9561 +
9562 +/* unionfs inode data in memory */
9563 +struct unionfs_inode_info {
9564 + int bstart;
9565 + int bend;
9566 + atomic_t generation;
9567 + /* Stuff for readdir over NFS. */
9568 + spinlock_t rdlock;
9569 + struct list_head readdircache;
9570 + int rdcount;
9571 + int hashsize;
9572 + int cookie;
9573 +
9574 + /* The lower inodes */
9575 + struct inode **lower_inodes;
9576 +
9577 + struct inode vfs_inode;
9578 +};
9579 +
9580 +/* unionfs dentry data in memory */
9581 +struct unionfs_dentry_info {
9582 + /*
9583 + * The semaphore is used to lock the dentry as soon as we get into a
9584 + * unionfs function from the VFS. Our lock ordering is that children
9585 + * go before their parents.
9586 + */
9587 + struct mutex lock;
9588 + int bstart;
9589 + int bend;
9590 + int bopaque;
9591 + int bcount;
9592 + atomic_t generation;
9593 + struct path *lower_paths;
9594 +};
9595 +
9596 +/* These are the pointers to our various objects. */
9597 +struct unionfs_data {
9598 + struct super_block *sb; /* lower super_block */
9599 + atomic_t open_files; /* number of open files on branch */
9600 + int branchperms;
9601 + int branch_id; /* unique branch ID at re/mount time */
9602 +};
9603 +
9604 +/* unionfs super-block data in memory */
9605 +struct unionfs_sb_info {
9606 + int bend;
9607 +
9608 + atomic_t generation;
9609 +
9610 + /*
9611 + * This rwsem is used to make sure that a branch management
9612 + * operation...
9613 + * 1) will not begin before all currently in-flight operations
9614 + * complete.
9615 + * 2) any new operations do not execute until the currently
9616 + * running branch management operation completes.
9617 + *
9618 + * The write_lock_owner records the PID of the task which grabbed
9619 + * the rw_sem for writing. If the same task also tries to grab the
9620 + * read lock, we allow it. This prevents a self-deadlock when
9621 + * branch-management is used on a pivot_root'ed union, because we
9622 + * have to ->lookup paths which belong to the same union.
9623 + */
9624 + struct unionfs_rw_semaphore rwsem;
9625 + pid_t write_lock_owner; /* PID of rw_sem owner (write lock) */
9626 + int high_branch_id; /* last unique branch ID given */
9627 + char *dev_name; /* to identify different unions in pr_debug */
9628 + struct unionfs_data *data;
9629 +};
9630 +
9631 +/*
9632 + * structure for making the linked list of entries by readdir on left branch
9633 + * to compare with entries on right branch
9634 + */
9635 +struct filldir_node {
9636 + struct list_head file_list; /* list for directory entries */
9637 + char *name; /* name entry */
9638 + int hash; /* name hash */
9639 + int namelen; /* name len since name is not 0 terminated */
9640 +
9641 + /*
9642 + * we can check for duplicate whiteouts and files in the same branch
9643 + * in order to return -EIO.
9644 + */
9645 + int bindex;
9646 +
9647 + /* is this a whiteout entry? */
9648 + int whiteout;
9649 +
9650 + /* Inline name, so we don't need to separately kmalloc small ones */
9651 + char iname[DNAME_INLINE_LEN_MIN];
9652 +};
9653 +
9654 +/* Directory hash table. */
9655 +struct unionfs_dir_state {
9656 + unsigned int cookie; /* the cookie, based off of rdversion */
9657 + unsigned int offset; /* The entry we have returned. */
9658 + int bindex;
9659 + loff_t dirpos; /* offset within the lower level directory */
9660 + int size; /* How big is the hash table? */
9661 + int hashentries; /* How many entries have been inserted? */
9662 + unsigned long access;
9663 +
9664 + /* This cache list is used when the inode keeps us around. */
9665 + struct list_head cache;
9666 + struct list_head list[0];
9667 +};
9668 +
9669 +/* externs needed for fanout.h or sioq.h */
9670 +extern int unionfs_get_nlinks(const struct inode *inode);
9671 +extern void unionfs_copy_attr_times(struct inode *upper);
9672 +extern void unionfs_copy_attr_all(struct inode *dest, const struct inode *src);
9673 +
9674 +/* include miscellaneous macros */
9675 +#include "fanout.h"
9676 +#include "sioq.h"
9677 +
9678 +/* externs for cache creation/deletion routines */
9679 +extern void unionfs_destroy_filldir_cache(void);
9680 +extern int unionfs_init_filldir_cache(void);
9681 +extern int unionfs_init_inode_cache(void);
9682 +extern void unionfs_destroy_inode_cache(void);
9683 +extern int unionfs_init_dentry_cache(void);
9684 +extern void unionfs_destroy_dentry_cache(void);
9685 +
9686 +/* Initialize and free readdir-specific state. */
9687 +extern int init_rdstate(struct file *file);
9688 +extern struct unionfs_dir_state *alloc_rdstate(struct inode *inode,
9689 + int bindex);
9690 +extern struct unionfs_dir_state *find_rdstate(struct inode *inode,
9691 + loff_t fpos);
9692 +extern void free_rdstate(struct unionfs_dir_state *state);
9693 +extern int add_filldir_node(struct unionfs_dir_state *rdstate,
9694 + const char *name, int namelen, int bindex,
9695 + int whiteout);
9696 +extern struct filldir_node *find_filldir_node(struct unionfs_dir_state *rdstate,
9697 + const char *name, int namelen,
9698 + int is_whiteout);
9699 +
9700 +extern struct dentry **alloc_new_dentries(int objs);
9701 +extern struct unionfs_data *alloc_new_data(int objs);
9702 +
9703 +/* We can only use 32-bits of offset for rdstate --- blech! */
9704 +#define DIREOF (0xfffff)
9705 +#define RDOFFBITS 20 /* This is the number of bits in DIREOF. */
9706 +#define MAXRDCOOKIE (0xfff)
9707 +/* Turn an rdstate into an offset. */
9708 +static inline off_t rdstate2offset(struct unionfs_dir_state *buf)
9709 +{
9710 + off_t tmp;
9711 +
9712 + tmp = ((buf->cookie & MAXRDCOOKIE) << RDOFFBITS)
9713 + | (buf->offset & DIREOF);
9714 + return tmp;
9715 +}
9716 +
9717 +/* Macros for locking a super_block. */
9718 +enum unionfs_super_lock_class {
9719 + UNIONFS_SMUTEX_NORMAL,
9720 + UNIONFS_SMUTEX_PARENT, /* when locking on behalf of file */
9721 + UNIONFS_SMUTEX_CHILD, /* when locking on behalf of dentry */
9722 +};
9723 +static inline void unionfs_read_lock(struct super_block *sb, int subclass)
9724 +{
9725 + if (UNIONFS_SB(sb)->write_lock_owner &&
9726 + UNIONFS_SB(sb)->write_lock_owner == current->pid)
9727 + return;
9728 + down_read_nested(&UNIONFS_SB(sb)->rwsem, subclass);
9729 +}
9730 +static inline void unionfs_read_unlock(struct super_block *sb)
9731 +{
9732 + if (UNIONFS_SB(sb)->write_lock_owner &&
9733 + UNIONFS_SB(sb)->write_lock_owner == current->pid)
9734 + return;
9735 + up_read(&UNIONFS_SB(sb)->rwsem);
9736 +}
9737 +static inline void unionfs_write_lock(struct super_block *sb)
9738 +{
9739 + down_write(&UNIONFS_SB(sb)->rwsem);
9740 + UNIONFS_SB(sb)->write_lock_owner = current->pid;
9741 +}
9742 +static inline void unionfs_write_unlock(struct super_block *sb)
9743 +{
9744 + up_write(&UNIONFS_SB(sb)->rwsem);
9745 + UNIONFS_SB(sb)->write_lock_owner = 0;
9746 +}
9747 +
9748 +static inline void unionfs_double_lock_dentry(struct dentry *d1,
9749 + struct dentry *d2)
9750 +{
9751 + BUG_ON(d1 == d2);
9752 + if (d1 < d2) {
9753 + unionfs_lock_dentry(d1, UNIONFS_DMUTEX_PARENT);
9754 + unionfs_lock_dentry(d2, UNIONFS_DMUTEX_CHILD);
9755 + } else {
9756 + unionfs_lock_dentry(d2, UNIONFS_DMUTEX_PARENT);
9757 + unionfs_lock_dentry(d1, UNIONFS_DMUTEX_CHILD);
9758 + }
9759 +}
9760 +
9761 +static inline void unionfs_double_unlock_dentry(struct dentry *d1,
9762 + struct dentry *d2)
9763 +{
9764 + BUG_ON(d1 == d2);
9765 + if (d1 < d2) { /* unlock in reverse order than double_lock_dentry */
9766 + unionfs_unlock_dentry(d1);
9767 + unionfs_unlock_dentry(d2);
9768 + } else {
9769 + unionfs_unlock_dentry(d2);
9770 + unionfs_unlock_dentry(d1);
9771 + }
9772 +}
9773 +
9774 +static inline void unionfs_double_lock_parents(struct dentry *p1,
9775 + struct dentry *p2)
9776 +{
9777 + if (p1 == p2) {
9778 + unionfs_lock_dentry(p1, UNIONFS_DMUTEX_REVAL_PARENT);
9779 + return;
9780 + }
9781 + if (p1 < p2) {
9782 + unionfs_lock_dentry(p1, UNIONFS_DMUTEX_REVAL_PARENT);
9783 + unionfs_lock_dentry(p2, UNIONFS_DMUTEX_REVAL_CHILD);
9784 + } else {
9785 + unionfs_lock_dentry(p2, UNIONFS_DMUTEX_REVAL_PARENT);
9786 + unionfs_lock_dentry(p1, UNIONFS_DMUTEX_REVAL_CHILD);
9787 + }
9788 +}
9789 +
9790 +static inline void unionfs_double_unlock_parents(struct dentry *p1,
9791 + struct dentry *p2)
9792 +{
9793 + if (p1 == p2) {
9794 + unionfs_unlock_dentry(p1);
9795 + return;
9796 + }
9797 + if (p1 < p2) { /* unlock in reverse order of double_lock_parents */
9798 + unionfs_unlock_dentry(p1);
9799 + unionfs_unlock_dentry(p2);
9800 + } else {
9801 + unionfs_unlock_dentry(p2);
9802 + unionfs_unlock_dentry(p1);
9803 + }
9804 +}
9805 +
9806 +extern int new_dentry_private_data(struct dentry *dentry, int subclass);
9807 +extern int realloc_dentry_private_data(struct dentry *dentry);
9808 +extern void free_dentry_private_data(struct dentry *dentry);
9809 +extern void update_bstart(struct dentry *dentry);
9810 +extern int init_lower_nd(struct nameidata *nd, unsigned int flags);
9811 +extern void release_lower_nd(struct nameidata *nd, int err);
9812 +
9813 +/*
9814 + * EXTERNALS:
9815 + */
9816 +
9817 +/* replicates the directory structure up to given dentry in given branch */
9818 +extern struct dentry *create_parents(struct inode *dir, struct dentry *dentry,
9819 + const char *name, int bindex);
9820 +
9821 +/* partial lookup */
9822 +extern int unionfs_partial_lookup(struct dentry *dentry,
9823 + struct dentry *parent);
9824 +extern struct dentry *unionfs_lookup_full(struct dentry *dentry,
9825 + struct dentry *parent,
9826 + int lookupmode);
9827 +
9828 +/* copies a file from dbstart to newbindex branch */
9829 +extern int copyup_file(struct inode *dir, struct file *file, int bstart,
9830 + int newbindex, loff_t size);
9831 +extern int copyup_named_file(struct inode *dir, struct file *file,
9832 + char *name, int bstart, int new_bindex,
9833 + loff_t len);
9834 +/* copies a dentry from dbstart to newbindex branch */
9835 +extern int copyup_dentry(struct inode *dir, struct dentry *dentry,
9836 + int bstart, int new_bindex, const char *name,
9837 + int namelen, struct file **copyup_file, loff_t len);
9838 +/* helper functions for post-copyup actions */
9839 +extern void unionfs_postcopyup_setmnt(struct dentry *dentry);
9840 +extern void unionfs_postcopyup_release(struct dentry *dentry);
9841 +
9842 +/* Is this directory empty: 0 if it is empty, -ENOTEMPTY if not. */
9843 +extern int check_empty(struct dentry *dentry, struct dentry *parent,
9844 + struct unionfs_dir_state **namelist);
9845 +/* whiteout and opaque directory helpers */
9846 +extern char *alloc_whname(const char *name, int len);
9847 +extern bool is_whiteout_name(char **namep, int *namelenp);
9848 +extern bool is_validname(const char *name);
9849 +extern struct dentry *lookup_whiteout(const char *name,
9850 + struct dentry *lower_parent);
9851 +extern struct dentry *find_first_whiteout(struct dentry *dentry);
9852 +extern int unlink_whiteout(struct dentry *wh_dentry);
9853 +extern int check_unlink_whiteout(struct dentry *dentry,
9854 + struct dentry *lower_dentry, int bindex);
9855 +extern int create_whiteout(struct dentry *dentry, int start);
9856 +extern int delete_whiteouts(struct dentry *dentry, int bindex,
9857 + struct unionfs_dir_state *namelist);
9858 +extern int is_opaque_dir(struct dentry *dentry, int bindex);
9859 +extern int make_dir_opaque(struct dentry *dir, int bindex);
9860 +extern void unionfs_set_max_namelen(long *namelen);
9861 +
9862 +extern void unionfs_reinterpose(struct dentry *this_dentry);
9863 +extern struct super_block *unionfs_duplicate_super(struct super_block *sb);
9864 +
9865 +/* Locking functions. */
9866 +extern int unionfs_setlk(struct file *file, int cmd, struct file_lock *fl);
9867 +extern int unionfs_getlk(struct file *file, struct file_lock *fl);
9868 +
9869 +/* Common file operations. */
9870 +extern int unionfs_file_revalidate(struct file *file, struct dentry *parent,
9871 + bool willwrite);
9872 +extern int unionfs_open(struct inode *inode, struct file *file);
9873 +extern int unionfs_file_release(struct inode *inode, struct file *file);
9874 +extern int unionfs_flush(struct file *file, fl_owner_t id);
9875 +extern long unionfs_ioctl(struct file *file, unsigned int cmd,
9876 + unsigned long arg);
9877 +extern int unionfs_fsync(struct file *file, struct dentry *dentry,
9878 + int datasync);
9879 +extern int unionfs_fasync(int fd, struct file *file, int flag);
9880 +
9881 +/* Inode operations */
9882 +extern struct inode *unionfs_iget(struct super_block *sb, unsigned long ino);
9883 +extern int unionfs_rename(struct inode *old_dir, struct dentry *old_dentry,
9884 + struct inode *new_dir, struct dentry *new_dentry);
9885 +extern int unionfs_unlink(struct inode *dir, struct dentry *dentry);
9886 +extern int unionfs_rmdir(struct inode *dir, struct dentry *dentry);
9887 +
9888 +extern bool __unionfs_d_revalidate(struct dentry *dentry,
9889 + struct dentry *parent, bool willwrite);
9890 +extern bool is_negative_lower(const struct dentry *dentry);
9891 +extern bool is_newer_lower(const struct dentry *dentry);
9892 +extern void purge_sb_data(struct super_block *sb);
9893 +
9894 +/* The values for unionfs_interpose's flag. */
9895 +#define INTERPOSE_DEFAULT 0
9896 +#define INTERPOSE_LOOKUP 1
9897 +#define INTERPOSE_REVAL 2
9898 +#define INTERPOSE_REVAL_NEG 3
9899 +#define INTERPOSE_PARTIAL 4
9900 +
9901 +extern struct dentry *unionfs_interpose(struct dentry *this_dentry,
9902 + struct super_block *sb, int flag);
9903 +
9904 +#ifdef CONFIG_UNION_FS_XATTR
9905 +/* Extended attribute functions. */
9906 +extern void *unionfs_xattr_alloc(size_t size, size_t limit);
9907 +static inline void unionfs_xattr_kfree(const void *p)
9908 +{
9909 + kfree(p);
9910 +}
9911 +extern ssize_t unionfs_getxattr(struct dentry *dentry, const char *name,
9912 + void *value, size_t size);
9913 +extern int unionfs_removexattr(struct dentry *dentry, const char *name);
9914 +extern ssize_t unionfs_listxattr(struct dentry *dentry, char *list,
9915 + size_t size);
9916 +extern int unionfs_setxattr(struct dentry *dentry, const char *name,
9917 + const void *value, size_t size, int flags);
9918 +#endif /* CONFIG_UNION_FS_XATTR */
9919 +
9920 +/* The root directory is unhashed, but isn't deleted. */
9921 +static inline int d_deleted(struct dentry *d)
9922 +{
9923 + return d_unhashed(d) && (d != d->d_sb->s_root);
9924 +}
9925 +
9926 +/* unionfs_permission, check if we should bypass error to facilitate copyup */
9927 +#define IS_COPYUP_ERR(err) ((err) == -EROFS)
9928 +
9929 +/* unionfs_open, check if we need to copyup the file */
9930 +#define OPEN_WRITE_FLAGS (O_WRONLY | O_RDWR | O_APPEND)
9931 +#define IS_WRITE_FLAG(flag) ((flag) & OPEN_WRITE_FLAGS)
9932 +
9933 +static inline int branchperms(const struct super_block *sb, int index)
9934 +{
9935 + BUG_ON(index < 0);
9936 + return UNIONFS_SB(sb)->data[index].branchperms;
9937 +}
9938 +
9939 +static inline int set_branchperms(struct super_block *sb, int index, int perms)
9940 +{
9941 + BUG_ON(index < 0);
9942 + UNIONFS_SB(sb)->data[index].branchperms = perms;
9943 + return perms;
9944 +}
9945 +
9946 +/* Is this file on a read-only branch? */
9947 +static inline int is_robranch_super(const struct super_block *sb, int index)
9948 +{
9949 + int ret;
9950 +
9951 + ret = (!(branchperms(sb, index) & MAY_WRITE)) ? -EROFS : 0;
9952 + return ret;
9953 +}
9954 +
9955 +/* Is this file on a read-only branch? */
9956 +static inline int is_robranch_idx(const struct dentry *dentry, int index)
9957 +{
9958 + struct super_block *lower_sb;
9959 +
9960 + BUG_ON(index < 0);
9961 +
9962 + if (!(branchperms(dentry->d_sb, index) & MAY_WRITE))
9963 + return -EROFS;
9964 +
9965 + lower_sb = unionfs_lower_super_idx(dentry->d_sb, index);
9966 + BUG_ON(lower_sb == NULL);
9967 + /*
9968 + * test sb flags directly, not IS_RDONLY(lower_inode) because the
9969 + * lower_dentry could be a negative.
9970 + */
9971 + if (lower_sb->s_flags & MS_RDONLY)
9972 + return -EROFS;
9973 +
9974 + return 0;
9975 +}
9976 +
9977 +static inline int is_robranch(const struct dentry *dentry)
9978 +{
9979 + int index;
9980 +
9981 + index = UNIONFS_D(dentry)->bstart;
9982 + BUG_ON(index < 0);
9983 +
9984 + return is_robranch_idx(dentry, index);
9985 +}
9986 +
9987 +/*
9988 + * EXTERNALS:
9989 + */
9990 +extern int check_branch(struct nameidata *nd);
9991 +extern int parse_branch_mode(const char *name, int *perms);
9992 +
9993 +/* locking helpers */
9994 +static inline struct dentry *lock_parent(struct dentry *dentry)
9995 +{
9996 + struct dentry *dir = dget_parent(dentry);
9997 + mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT);
9998 + return dir;
9999 +}
10000 +static inline struct dentry *lock_parent_wh(struct dentry *dentry)
10001 +{
10002 + struct dentry *dir = dget_parent(dentry);
10003 +
10004 + mutex_lock_nested(&dir->d_inode->i_mutex, UNIONFS_DMUTEX_WHITEOUT);
10005 + return dir;
10006 +}
10007 +
10008 +static inline void unlock_dir(struct dentry *dir)
10009 +{
10010 + mutex_unlock(&dir->d_inode->i_mutex);
10011 + dput(dir);
10012 +}
10013 +
10014 +static inline struct vfsmount *unionfs_mntget(struct dentry *dentry,
10015 + int bindex)
10016 +{
10017 + struct vfsmount *mnt;
10018 +
10019 + BUG_ON(!dentry || bindex < 0);
10020 +
10021 + mnt = mntget(unionfs_lower_mnt_idx(dentry, bindex));
10022 +#ifdef CONFIG_UNION_FS_DEBUG
10023 + if (!mnt)
10024 + pr_debug("unionfs: mntget: mnt=%p bindex=%d\n",
10025 + mnt, bindex);
10026 +#endif /* CONFIG_UNION_FS_DEBUG */
10027 +
10028 + return mnt;
10029 +}
10030 +
10031 +static inline void unionfs_mntput(struct dentry *dentry, int bindex)
10032 +{
10033 + struct vfsmount *mnt;
10034 +
10035 + if (!dentry && bindex < 0)
10036 + return;
10037 + BUG_ON(!dentry || bindex < 0);
10038 +
10039 + mnt = unionfs_lower_mnt_idx(dentry, bindex);
10040 +#ifdef CONFIG_UNION_FS_DEBUG
10041 + /*
10042 + * Directories can have NULL lower objects in between start/end, but
10043 + * NOT if at the start/end range. We cannot verify that this dentry
10044 + * is a type=DIR, because it may already be a negative dentry. But
10045 + * if dbstart is greater than dbend, we know that this couldn't have
10046 + * been a regular file: it had to have been a directory.
10047 + */
10048 + if (!mnt && !(bindex > dbstart(dentry) && bindex < dbend(dentry)))
10049 + pr_debug("unionfs: mntput: mnt=%p bindex=%d\n", mnt, bindex);
10050 +#endif /* CONFIG_UNION_FS_DEBUG */
10051 + mntput(mnt);
10052 +}
10053 +
10054 +#ifdef CONFIG_UNION_FS_DEBUG
10055 +
10056 +/* useful for tracking code reachability */
10057 +#define UDBG pr_debug("DBG:%s:%s:%d\n", __FILE__, __func__, __LINE__)
10058 +
10059 +#define unionfs_check_inode(i) __unionfs_check_inode((i), \
10060 + __FILE__, __func__, __LINE__)
10061 +#define unionfs_check_dentry(d) __unionfs_check_dentry((d), \
10062 + __FILE__, __func__, __LINE__)
10063 +#define unionfs_check_file(f) __unionfs_check_file((f), \
10064 + __FILE__, __func__, __LINE__)
10065 +#define unionfs_check_nd(n) __unionfs_check_nd((n), \
10066 + __FILE__, __func__, __LINE__)
10067 +#define show_branch_counts(sb) __show_branch_counts((sb), \
10068 + __FILE__, __func__, __LINE__)
10069 +#define show_inode_times(i) __show_inode_times((i), \
10070 + __FILE__, __func__, __LINE__)
10071 +#define show_dinode_times(d) __show_dinode_times((d), \
10072 + __FILE__, __func__, __LINE__)
10073 +#define show_inode_counts(i) __show_inode_counts((i), \
10074 + __FILE__, __func__, __LINE__)
10075 +
10076 +extern void __unionfs_check_inode(const struct inode *inode, const char *fname,
10077 + const char *fxn, int line);
10078 +extern void __unionfs_check_dentry(const struct dentry *dentry,
10079 + const char *fname, const char *fxn,
10080 + int line);
10081 +extern void __unionfs_check_file(const struct file *file,
10082 + const char *fname, const char *fxn, int line);
10083 +extern void __unionfs_check_nd(const struct nameidata *nd,
10084 + const char *fname, const char *fxn, int line);
10085 +extern void __show_branch_counts(const struct super_block *sb,
10086 + const char *file, const char *fxn, int line);
10087 +extern void __show_inode_times(const struct inode *inode,
10088 + const char *file, const char *fxn, int line);
10089 +extern void __show_dinode_times(const struct dentry *dentry,
10090 + const char *file, const char *fxn, int line);
10091 +extern void __show_inode_counts(const struct inode *inode,
10092 + const char *file, const char *fxn, int line);
10093 +
10094 +#else /* not CONFIG_UNION_FS_DEBUG */
10095 +
10096 +/* we leave useful hooks for these check functions throughout the code */
10097 +#define unionfs_check_inode(i) do { } while (0)
10098 +#define unionfs_check_dentry(d) do { } while (0)
10099 +#define unionfs_check_file(f) do { } while (0)
10100 +#define unionfs_check_nd(n) do { } while (0)
10101 +#define show_branch_counts(sb) do { } while (0)
10102 +#define show_inode_times(i) do { } while (0)
10103 +#define show_dinode_times(d) do { } while (0)
10104 +#define show_inode_counts(i) do { } while (0)
10105 +
10106 +#endif /* not CONFIG_UNION_FS_DEBUG */
10107 +
10108 +#endif /* not _UNION_H_ */
10109 diff -Naur linux-2.6.30/fs/unionfs/unlink.c linux-2.6.30-magellan/fs/unionfs/unlink.c
10110 --- linux-2.6.30/fs/unionfs/unlink.c 1970-01-01 01:00:00.000000000 +0100
10111 +++ linux-2.6.30-magellan/fs/unionfs/unlink.c 2009-06-11 18:39:19.000000000 +0200
10112 @@ -0,0 +1,282 @@
10113 +/*
10114 + * Copyright (c) 2003-2009 Erez Zadok
10115 + * Copyright (c) 2003-2006 Charles P. Wright
10116 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
10117 + * Copyright (c) 2005-2006 Junjiro Okajima
10118 + * Copyright (c) 2005 Arun M. Krishnakumar
10119 + * Copyright (c) 2004-2006 David P. Quigley
10120 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
10121 + * Copyright (c) 2003 Puja Gupta
10122 + * Copyright (c) 2003 Harikesavan Krishnan
10123 + * Copyright (c) 2003-2009 Stony Brook University
10124 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
10125 + *
10126 + * This program is free software; you can redistribute it and/or modify
10127 + * it under the terms of the GNU General Public License version 2 as
10128 + * published by the Free Software Foundation.
10129 + */
10130 +
10131 +#include "union.h"
10132 +
10133 +/*
10134 + * Helper function for Unionfs's unlink operation.
10135 + *
10136 + * The main goal of this function is to optimize the unlinking of non-dir
10137 + * objects in unionfs by deleting all possible lower inode objects from the
10138 + * underlying branches having same dentry name as the non-dir dentry on
10139 + * which this unlink operation is called. This way we delete as many lower
10140 + * inodes as possible, and save space. Whiteouts need to be created in
10141 + * branch0 only if unlinking fails on any of the lower branch other than
10142 + * branch0, or if a lower branch is marked read-only.
10143 + *
10144 + * Also, while unlinking a file, if we encounter any dir type entry in any
10145 + * intermediate branch, then we remove the directory by calling vfs_rmdir.
10146 + * The following special cases are also handled:
10147 +
10148 + * (1) If an error occurs in branch0 during vfs_unlink, then we return
10149 + * appropriate error.
10150 + *
10151 + * (2) If we get an error during unlink in any of other lower branch other
10152 + * than branch0, then we create a whiteout in branch0.
10153 + *
10154 + * (3) If a whiteout already exists in any intermediate branch, we delete
10155 + * all possible inodes only up to that branch (this is an "opaqueness"
10156 + * as as per Documentation/filesystems/unionfs/concepts.txt).
10157 + *
10158 + */
10159 +static int unionfs_unlink_whiteout(struct inode *dir, struct dentry *dentry,
10160 + struct dentry *parent)
10161 +{
10162 + struct dentry *lower_dentry;
10163 + struct dentry *lower_dir_dentry;
10164 + int bindex;
10165 + int err = 0;
10166 +
10167 + err = unionfs_partial_lookup(dentry, parent);
10168 + if (err)
10169 + goto out;
10170 +
10171 + /* trying to unlink all possible valid instances */
10172 + for (bindex = dbstart(dentry); bindex <= dbend(dentry); bindex++) {
10173 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
10174 + if (!lower_dentry || !lower_dentry->d_inode)
10175 + continue;
10176 +
10177 + lower_dir_dentry = lock_parent(lower_dentry);
10178 +
10179 + /* avoid destroying the lower inode if the object is in use */
10180 + dget(lower_dentry);
10181 + err = is_robranch_super(dentry->d_sb, bindex);
10182 + if (!err) {
10183 + /* see Documentation/filesystems/unionfs/issues.txt */
10184 + lockdep_off();
10185 + if (!S_ISDIR(lower_dentry->d_inode->i_mode))
10186 + err = vfs_unlink(lower_dir_dentry->d_inode,
10187 + lower_dentry);
10188 + else
10189 + err = vfs_rmdir(lower_dir_dentry->d_inode,
10190 + lower_dentry);
10191 + lockdep_on();
10192 + }
10193 +
10194 + /* if lower object deletion succeeds, update inode's times */
10195 + if (!err)
10196 + unionfs_copy_attr_times(dentry->d_inode);
10197 + dput(lower_dentry);
10198 + fsstack_copy_attr_times(dir, lower_dir_dentry->d_inode);
10199 + unlock_dir(lower_dir_dentry);
10200 +
10201 + if (err)
10202 + break;
10203 + }
10204 +
10205 + /*
10206 + * Create the whiteout in branch 0 (highest priority) only if (a)
10207 + * there was an error in any intermediate branch other than branch 0
10208 + * due to failure of vfs_unlink/vfs_rmdir or (b) a branch marked or
10209 + * mounted read-only.
10210 + */
10211 + if (err) {
10212 + if ((bindex == 0) ||
10213 + ((bindex == dbstart(dentry)) &&
10214 + (!IS_COPYUP_ERR(err))))
10215 + goto out;
10216 + else {
10217 + if (!IS_COPYUP_ERR(err))
10218 + pr_debug("unionfs: lower object deletion "
10219 + "failed in branch:%d\n", bindex);
10220 + err = create_whiteout(dentry, sbstart(dentry->d_sb));
10221 + }
10222 + }
10223 +
10224 +out:
10225 + if (!err)
10226 + inode_dec_link_count(dentry->d_inode);
10227 +
10228 + /* We don't want to leave negative leftover dentries for revalidate. */
10229 + if (!err && (dbopaque(dentry) != -1))
10230 + update_bstart(dentry);
10231 +
10232 + return err;
10233 +}
10234 +
10235 +int unionfs_unlink(struct inode *dir, struct dentry *dentry)
10236 +{
10237 + int err = 0;
10238 + struct inode *inode = dentry->d_inode;
10239 + struct dentry *parent;
10240 + int valid;
10241 +
10242 + BUG_ON(S_ISDIR(inode->i_mode));
10243 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
10244 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
10245 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
10246 +
10247 + valid = __unionfs_d_revalidate(dentry, parent, false);
10248 + if (unlikely(!valid)) {
10249 + err = -ESTALE;
10250 + goto out;
10251 + }
10252 + unionfs_check_dentry(dentry);
10253 +
10254 + err = unionfs_unlink_whiteout(dir, dentry, parent);
10255 + /* call d_drop so the system "forgets" about us */
10256 + if (!err) {
10257 + unionfs_postcopyup_release(dentry);
10258 + unionfs_postcopyup_setmnt(parent);
10259 + if (inode->i_nlink == 0) /* drop lower inodes */
10260 + iput_lowers_all(inode, false);
10261 + d_drop(dentry);
10262 + /*
10263 + * if unlink/whiteout succeeded, parent dir mtime has
10264 + * changed
10265 + */
10266 + unionfs_copy_attr_times(dir);
10267 + }
10268 +
10269 +out:
10270 + if (!err) {
10271 + unionfs_check_dentry(dentry);
10272 + unionfs_check_inode(dir);
10273 + }
10274 + unionfs_unlock_dentry(dentry);
10275 + unionfs_unlock_parent(dentry, parent);
10276 + unionfs_read_unlock(dentry->d_sb);
10277 + return err;
10278 +}
10279 +
10280 +static int unionfs_rmdir_first(struct inode *dir, struct dentry *dentry,
10281 + struct unionfs_dir_state *namelist)
10282 +{
10283 + int err;
10284 + struct dentry *lower_dentry;
10285 + struct dentry *lower_dir_dentry = NULL;
10286 +
10287 + /* Here we need to remove whiteout entries. */
10288 + err = delete_whiteouts(dentry, dbstart(dentry), namelist);
10289 + if (err)
10290 + goto out;
10291 +
10292 + lower_dentry = unionfs_lower_dentry(dentry);
10293 +
10294 + lower_dir_dentry = lock_parent(lower_dentry);
10295 +
10296 + /* avoid destroying the lower inode if the file is in use */
10297 + dget(lower_dentry);
10298 + err = is_robranch(dentry);
10299 + if (!err) {
10300 + /* see Documentation/filesystems/unionfs/issues.txt */
10301 + lockdep_off();
10302 + err = vfs_rmdir(lower_dir_dentry->d_inode, lower_dentry);
10303 + lockdep_on();
10304 + }
10305 + dput(lower_dentry);
10306 +
10307 + fsstack_copy_attr_times(dir, lower_dir_dentry->d_inode);
10308 + /* propagate number of hard-links */
10309 + dentry->d_inode->i_nlink = unionfs_get_nlinks(dentry->d_inode);
10310 +
10311 +out:
10312 + if (lower_dir_dentry)
10313 + unlock_dir(lower_dir_dentry);
10314 + return err;
10315 +}
10316 +
10317 +int unionfs_rmdir(struct inode *dir, struct dentry *dentry)
10318 +{
10319 + int err = 0;
10320 + struct unionfs_dir_state *namelist = NULL;
10321 + struct dentry *parent;
10322 + int dstart, dend;
10323 + bool valid;
10324 +
10325 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
10326 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
10327 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
10328 +
10329 + valid = __unionfs_d_revalidate(dentry, parent, false);
10330 + if (unlikely(!valid)) {
10331 + err = -ESTALE;
10332 + goto out;
10333 + }
10334 + unionfs_check_dentry(dentry);
10335 +
10336 + /* check if this unionfs directory is empty or not */
10337 + err = check_empty(dentry, parent, &namelist);
10338 + if (err)
10339 + goto out;
10340 +
10341 + err = unionfs_rmdir_first(dir, dentry, namelist);
10342 + dstart = dbstart(dentry);
10343 + dend = dbend(dentry);
10344 + /*
10345 + * We create a whiteout for the directory if there was an error to
10346 + * rmdir the first directory entry in the union. Otherwise, we
10347 + * create a whiteout only if there is no chance that a lower
10348 + * priority branch might also have the same named directory. IOW,
10349 + * if there is not another same-named directory at a lower priority
10350 + * branch, then we don't need to create a whiteout for it.
10351 + */
10352 + if (!err) {
10353 + if (dstart < dend)
10354 + err = create_whiteout(dentry, dstart);
10355 + } else {
10356 + int new_err;
10357 +
10358 + if (dstart == 0)
10359 + goto out;
10360 +
10361 + /* exit if the error returned was NOT -EROFS */
10362 + if (!IS_COPYUP_ERR(err))
10363 + goto out;
10364 +
10365 + new_err = create_whiteout(dentry, dstart - 1);
10366 + if (new_err != -EEXIST)
10367 + err = new_err;
10368 + }
10369 +
10370 +out:
10371 + /*
10372 + * Drop references to lower dentry/inode so storage space for them
10373 + * can be reclaimed. Then, call d_drop so the system "forgets"
10374 + * about us.
10375 + */
10376 + if (!err) {
10377 + iput_lowers_all(dentry->d_inode, false);
10378 + dput(unionfs_lower_dentry_idx(dentry, dstart));
10379 + unionfs_set_lower_dentry_idx(dentry, dstart, NULL);
10380 + d_drop(dentry);
10381 + /* update our lower vfsmnts, in case a copyup took place */
10382 + unionfs_postcopyup_setmnt(dentry);
10383 + unionfs_check_dentry(dentry);
10384 + unionfs_check_inode(dir);
10385 + }
10386 +
10387 + if (namelist)
10388 + free_rdstate(namelist);
10389 +
10390 + unionfs_unlock_dentry(dentry);
10391 + unionfs_unlock_parent(dentry, parent);
10392 + unionfs_read_unlock(dentry->d_sb);
10393 + return err;
10394 +}
10395 diff -Naur linux-2.6.30/fs/unionfs/whiteout.c linux-2.6.30-magellan/fs/unionfs/whiteout.c
10396 --- linux-2.6.30/fs/unionfs/whiteout.c 1970-01-01 01:00:00.000000000 +0100
10397 +++ linux-2.6.30-magellan/fs/unionfs/whiteout.c 2009-06-11 18:39:19.000000000 +0200
10398 @@ -0,0 +1,584 @@
10399 +/*
10400 + * Copyright (c) 2003-2009 Erez Zadok
10401 + * Copyright (c) 2003-2006 Charles P. Wright
10402 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
10403 + * Copyright (c) 2005-2006 Junjiro Okajima
10404 + * Copyright (c) 2005 Arun M. Krishnakumar
10405 + * Copyright (c) 2004-2006 David P. Quigley
10406 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
10407 + * Copyright (c) 2003 Puja Gupta
10408 + * Copyright (c) 2003 Harikesavan Krishnan
10409 + * Copyright (c) 2003-2009 Stony Brook University
10410 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
10411 + *
10412 + * This program is free software; you can redistribute it and/or modify
10413 + * it under the terms of the GNU General Public License version 2 as
10414 + * published by the Free Software Foundation.
10415 + */
10416 +
10417 +#include "union.h"
10418 +
10419 +/*
10420 + * whiteout and opaque directory helpers
10421 + */
10422 +
10423 +/* What do we use for whiteouts. */
10424 +#define UNIONFS_WHPFX ".wh."
10425 +#define UNIONFS_WHLEN 4
10426 +/*
10427 + * If a directory contains this file, then it is opaque. We start with the
10428 + * .wh. flag so that it is blocked by lookup.
10429 + */
10430 +#define UNIONFS_DIR_OPAQUE_NAME "__dir_opaque"
10431 +#define UNIONFS_DIR_OPAQUE UNIONFS_WHPFX UNIONFS_DIR_OPAQUE_NAME
10432 +
10433 +/* construct whiteout filename */
10434 +char *alloc_whname(const char *name, int len)
10435 +{
10436 + char *buf;
10437 +
10438 + buf = kmalloc(len + UNIONFS_WHLEN + 1, GFP_KERNEL);
10439 + if (unlikely(!buf))
10440 + return ERR_PTR(-ENOMEM);
10441 +
10442 + strcpy(buf, UNIONFS_WHPFX);
10443 + strlcat(buf, name, len + UNIONFS_WHLEN + 1);
10444 +
10445 + return buf;
10446 +}
10447 +
10448 +/*
10449 + * XXX: this can be inline or CPP macro, but is here to keep all whiteout
10450 + * code in one place.
10451 + */
10452 +void unionfs_set_max_namelen(long *namelen)
10453 +{
10454 + *namelen -= UNIONFS_WHLEN;
10455 +}
10456 +
10457 +/* check if @namep is a whiteout, update @namep and @namelenp accordingly */
10458 +bool is_whiteout_name(char **namep, int *namelenp)
10459 +{
10460 + if (*namelenp > UNIONFS_WHLEN &&
10461 + !strncmp(*namep, UNIONFS_WHPFX, UNIONFS_WHLEN)) {
10462 + *namep += UNIONFS_WHLEN;
10463 + *namelenp -= UNIONFS_WHLEN;
10464 + return true;
10465 + }
10466 + return false;
10467 +}
10468 +
10469 +/* is the filename valid == !(whiteout for a file or opaque dir marker) */
10470 +bool is_validname(const char *name)
10471 +{
10472 + if (!strncmp(name, UNIONFS_WHPFX, UNIONFS_WHLEN))
10473 + return false;
10474 + if (!strncmp(name, UNIONFS_DIR_OPAQUE_NAME,
10475 + sizeof(UNIONFS_DIR_OPAQUE_NAME) - 1))
10476 + return false;
10477 + return true;
10478 +}
10479 +
10480 +/*
10481 + * Look for a whiteout @name in @lower_parent directory. If error, return
10482 + * ERR_PTR. Caller must dput() the returned dentry if not an error.
10483 + *
10484 + * XXX: some callers can reuse the whname allocated buffer to avoid repeated
10485 + * free then re-malloc calls. Need to provide a different API for those
10486 + * callers.
10487 + */
10488 +struct dentry *lookup_whiteout(const char *name, struct dentry *lower_parent)
10489 +{
10490 + char *whname = NULL;
10491 + int err = 0, namelen;
10492 + struct dentry *wh_dentry = NULL;
10493 +
10494 + namelen = strlen(name);
10495 + whname = alloc_whname(name, namelen);
10496 + if (unlikely(IS_ERR(whname))) {
10497 + err = PTR_ERR(whname);
10498 + goto out;
10499 + }
10500 +
10501 + /* check if whiteout exists in this branch: lookup .wh.foo */
10502 + wh_dentry = lookup_one_len(whname, lower_parent, strlen(whname));
10503 + if (IS_ERR(wh_dentry)) {
10504 + err = PTR_ERR(wh_dentry);
10505 + goto out;
10506 + }
10507 +
10508 + /* check if negative dentry (ENOENT) */
10509 + if (!wh_dentry->d_inode)
10510 + goto out;
10511 +
10512 + /* whiteout found: check if valid type */
10513 + if (!S_ISREG(wh_dentry->d_inode->i_mode)) {
10514 + printk(KERN_ERR "unionfs: invalid whiteout %s entry type %d\n",
10515 + whname, wh_dentry->d_inode->i_mode);
10516 + dput(wh_dentry);
10517 + err = -EIO;
10518 + goto out;
10519 + }
10520 +
10521 +out:
10522 + kfree(whname);
10523 + if (err)
10524 + wh_dentry = ERR_PTR(err);
10525 + return wh_dentry;
10526 +}
10527 +
10528 +/* find and return first whiteout in parent directory, else ENOENT */
10529 +struct dentry *find_first_whiteout(struct dentry *dentry)
10530 +{
10531 + int bindex, bstart, bend;
10532 + struct dentry *parent, *lower_parent, *wh_dentry;
10533 +
10534 + parent = dget_parent(dentry);
10535 +
10536 + bstart = dbstart(parent);
10537 + bend = dbend(parent);
10538 + wh_dentry = ERR_PTR(-ENOENT);
10539 +
10540 + for (bindex = bstart; bindex <= bend; bindex++) {
10541 + lower_parent = unionfs_lower_dentry_idx(parent, bindex);
10542 + if (!lower_parent)
10543 + continue;
10544 + wh_dentry = lookup_whiteout(dentry->d_name.name, lower_parent);
10545 + if (IS_ERR(wh_dentry))
10546 + continue;
10547 + if (wh_dentry->d_inode)
10548 + break;
10549 + dput(wh_dentry);
10550 + wh_dentry = ERR_PTR(-ENOENT);
10551 + }
10552 +
10553 + dput(parent);
10554 +
10555 + return wh_dentry;
10556 +}
10557 +
10558 +/*
10559 + * Unlink a whiteout dentry. Returns 0 or -errno. Caller must hold and
10560 + * release dentry reference.
10561 + */
10562 +int unlink_whiteout(struct dentry *wh_dentry)
10563 +{
10564 + int err;
10565 + struct dentry *lower_dir_dentry;
10566 +
10567 + /* dget and lock parent dentry */
10568 + lower_dir_dentry = lock_parent_wh(wh_dentry);
10569 +
10570 + /* see Documentation/filesystems/unionfs/issues.txt */
10571 + lockdep_off();
10572 + err = vfs_unlink(lower_dir_dentry->d_inode, wh_dentry);
10573 + lockdep_on();
10574 + unlock_dir(lower_dir_dentry);
10575 +
10576 + /*
10577 + * Whiteouts are special files and should be deleted no matter what
10578 + * (as if they never existed), in order to allow this create
10579 + * operation to succeed. This is especially important in sticky
10580 + * directories: a whiteout may have been created by one user, but
10581 + * the newly created file may be created by another user.
10582 + * Therefore, in order to maintain Unix semantics, if the vfs_unlink
10583 + * above failed, then we have to try to directly unlink the
10584 + * whiteout. Note: in the ODF version of unionfs, whiteout are
10585 + * handled much more cleanly.
10586 + */
10587 + if (err == -EPERM) {
10588 + struct inode *inode = lower_dir_dentry->d_inode;
10589 + err = inode->i_op->unlink(inode, wh_dentry);
10590 + }
10591 + if (err)
10592 + printk(KERN_ERR "unionfs: could not unlink whiteout %s, "
10593 + "err = %d\n", wh_dentry->d_name.name, err);
10594 +
10595 + return err;
10596 +
10597 +}
10598 +
10599 +/*
10600 + * Helper function when creating new objects (create, symlink, mknod, etc.).
10601 + * Checks to see if there's a whiteout in @lower_dentry's parent directory,
10602 + * whose name is taken from @dentry. Then tries to remove that whiteout, if
10603 + * found. If <dentry,bindex> is a branch marked readonly, return -EROFS.
10604 + * If it finds both a regular file and a whiteout, return -EIO (this should
10605 + * never happen).
10606 + *
10607 + * Return 0 if no whiteout was found. Return 1 if one was found and
10608 + * successfully removed. Therefore a value >= 0 tells the caller that
10609 + * @lower_dentry belongs to a good branch to create the new object in).
10610 + * Return -ERRNO if an error occurred during whiteout lookup or in trying to
10611 + * unlink the whiteout.
10612 + */
10613 +int check_unlink_whiteout(struct dentry *dentry, struct dentry *lower_dentry,
10614 + int bindex)
10615 +{
10616 + int err;
10617 + struct dentry *wh_dentry = NULL;
10618 + struct dentry *lower_dir_dentry = NULL;
10619 +
10620 + /* look for whiteout dentry first */
10621 + lower_dir_dentry = dget_parent(lower_dentry);
10622 + wh_dentry = lookup_whiteout(dentry->d_name.name, lower_dir_dentry);
10623 + dput(lower_dir_dentry);
10624 + if (IS_ERR(wh_dentry)) {
10625 + err = PTR_ERR(wh_dentry);
10626 + goto out;
10627 + }
10628 +
10629 + if (!wh_dentry->d_inode) { /* no whiteout exists*/
10630 + err = 0;
10631 + goto out_dput;
10632 + }
10633 +
10634 + /* check if regular file and whiteout were both found */
10635 + if (unlikely(lower_dentry->d_inode)) {
10636 + err = -EIO;
10637 + printk(KERN_ERR "unionfs: found both whiteout and regular "
10638 + "file in directory %s (branch %d)\n",
10639 + lower_dir_dentry->d_name.name, bindex);
10640 + goto out_dput;
10641 + }
10642 +
10643 + /* check if branch is writeable */
10644 + err = is_robranch_super(dentry->d_sb, bindex);
10645 + if (err)
10646 + goto out_dput;
10647 +
10648 + /* .wh.foo has been found, so let's unlink it */
10649 + err = unlink_whiteout(wh_dentry);
10650 + if (!err)
10651 + err = 1; /* a whiteout was found and successfully removed */
10652 +out_dput:
10653 + dput(wh_dentry);
10654 +out:
10655 + return err;
10656 +}
10657 +
10658 +/*
10659 + * Pass an unionfs dentry and an index. It will try to create a whiteout
10660 + * for the filename in dentry, and will try in branch 'index'. On error,
10661 + * it will proceed to a branch to the left.
10662 + */
10663 +int create_whiteout(struct dentry *dentry, int start)
10664 +{
10665 + int bstart, bend, bindex;
10666 + struct dentry *lower_dir_dentry;
10667 + struct dentry *lower_dentry;
10668 + struct dentry *lower_wh_dentry;
10669 + struct nameidata nd;
10670 + char *name = NULL;
10671 + int err = -EINVAL;
10672 +
10673 + verify_locked(dentry);
10674 +
10675 + bstart = dbstart(dentry);
10676 + bend = dbend(dentry);
10677 +
10678 + /* create dentry's whiteout equivalent */
10679 + name = alloc_whname(dentry->d_name.name, dentry->d_name.len);
10680 + if (unlikely(IS_ERR(name))) {
10681 + err = PTR_ERR(name);
10682 + goto out;
10683 + }
10684 +
10685 + for (bindex = start; bindex >= 0; bindex--) {
10686 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
10687 +
10688 + if (!lower_dentry) {
10689 + /*
10690 + * if lower dentry is not present, create the
10691 + * entire lower dentry directory structure and go
10692 + * ahead. Since we want to just create whiteout, we
10693 + * only want the parent dentry, and hence get rid of
10694 + * this dentry.
10695 + */
10696 + lower_dentry = create_parents(dentry->d_inode,
10697 + dentry,
10698 + dentry->d_name.name,
10699 + bindex);
10700 + if (!lower_dentry || IS_ERR(lower_dentry)) {
10701 + int ret = PTR_ERR(lower_dentry);
10702 + if (!IS_COPYUP_ERR(ret))
10703 + printk(KERN_ERR
10704 + "unionfs: create_parents for "
10705 + "whiteout failed: bindex=%d "
10706 + "err=%d\n", bindex, ret);
10707 + continue;
10708 + }
10709 + }
10710 +
10711 + lower_wh_dentry =
10712 + lookup_one_len(name, lower_dentry->d_parent,
10713 + dentry->d_name.len + UNIONFS_WHLEN);
10714 + if (IS_ERR(lower_wh_dentry))
10715 + continue;
10716 +
10717 + /*
10718 + * The whiteout already exists. This used to be impossible,
10719 + * but now is possible because of opaqueness.
10720 + */
10721 + if (lower_wh_dentry->d_inode) {
10722 + dput(lower_wh_dentry);
10723 + err = 0;
10724 + goto out;
10725 + }
10726 +
10727 + err = init_lower_nd(&nd, LOOKUP_CREATE);
10728 + if (unlikely(err < 0))
10729 + goto out;
10730 + lower_dir_dentry = lock_parent_wh(lower_wh_dentry);
10731 + err = is_robranch_super(dentry->d_sb, bindex);
10732 + if (!err)
10733 + err = vfs_create(lower_dir_dentry->d_inode,
10734 + lower_wh_dentry,
10735 + ~current->fs->umask & S_IRUGO,
10736 + &nd);
10737 + unlock_dir(lower_dir_dentry);
10738 + dput(lower_wh_dentry);
10739 + release_lower_nd(&nd, err);
10740 +
10741 + if (!err || !IS_COPYUP_ERR(err))
10742 + break;
10743 + }
10744 +
10745 + /* set dbopaque so that lookup will not proceed after this branch */
10746 + if (!err)
10747 + dbopaque(dentry) = bindex;
10748 +
10749 +out:
10750 + kfree(name);
10751 + return err;
10752 +}
10753 +
10754 +/*
10755 + * Delete all of the whiteouts in a given directory for rmdir.
10756 + *
10757 + * lower directory inode should be locked
10758 + */
10759 +static int do_delete_whiteouts(struct dentry *dentry, int bindex,
10760 + struct unionfs_dir_state *namelist)
10761 +{
10762 + int err = 0;
10763 + struct dentry *lower_dir_dentry = NULL;
10764 + struct dentry *lower_dentry;
10765 + char *name = NULL, *p;
10766 + struct inode *lower_dir;
10767 + int i;
10768 + struct list_head *pos;
10769 + struct filldir_node *cursor;
10770 +
10771 + /* Find out lower parent dentry */
10772 + lower_dir_dentry = unionfs_lower_dentry_idx(dentry, bindex);
10773 + BUG_ON(!S_ISDIR(lower_dir_dentry->d_inode->i_mode));
10774 + lower_dir = lower_dir_dentry->d_inode;
10775 + BUG_ON(!S_ISDIR(lower_dir->i_mode));
10776 +
10777 + err = -ENOMEM;
10778 + name = __getname();
10779 + if (unlikely(!name))
10780 + goto out;
10781 + strcpy(name, UNIONFS_WHPFX);
10782 + p = name + UNIONFS_WHLEN;
10783 +
10784 + err = 0;
10785 + for (i = 0; !err && i < namelist->size; i++) {
10786 + list_for_each(pos, &namelist->list[i]) {
10787 + cursor =
10788 + list_entry(pos, struct filldir_node,
10789 + file_list);
10790 + /* Only operate on whiteouts in this branch. */
10791 + if (cursor->bindex != bindex)
10792 + continue;
10793 + if (!cursor->whiteout)
10794 + continue;
10795 +
10796 + strlcpy(p, cursor->name, PATH_MAX - UNIONFS_WHLEN);
10797 + lower_dentry =
10798 + lookup_one_len(name, lower_dir_dentry,
10799 + cursor->namelen +
10800 + UNIONFS_WHLEN);
10801 + if (IS_ERR(lower_dentry)) {
10802 + err = PTR_ERR(lower_dentry);
10803 + break;
10804 + }
10805 + if (lower_dentry->d_inode)
10806 + err = vfs_unlink(lower_dir, lower_dentry);
10807 + dput(lower_dentry);
10808 + if (err)
10809 + break;
10810 + }
10811 + }
10812 +
10813 + __putname(name);
10814 +
10815 + /* After all of the removals, we should copy the attributes once. */
10816 + fsstack_copy_attr_times(dentry->d_inode, lower_dir_dentry->d_inode);
10817 +
10818 +out:
10819 + return err;
10820 +}
10821 +
10822 +
10823 +void __delete_whiteouts(struct work_struct *work)
10824 +{
10825 + struct sioq_args *args = container_of(work, struct sioq_args, work);
10826 + struct deletewh_args *d = &args->deletewh;
10827 +
10828 + args->err = do_delete_whiteouts(d->dentry, d->bindex, d->namelist);
10829 + complete(&args->comp);
10830 +}
10831 +
10832 +/* delete whiteouts in a dir (for rmdir operation) using sioq if necessary */
10833 +int delete_whiteouts(struct dentry *dentry, int bindex,
10834 + struct unionfs_dir_state *namelist)
10835 +{
10836 + int err;
10837 + struct super_block *sb;
10838 + struct dentry *lower_dir_dentry;
10839 + struct inode *lower_dir;
10840 + struct sioq_args args;
10841 +
10842 + sb = dentry->d_sb;
10843 +
10844 + BUG_ON(!S_ISDIR(dentry->d_inode->i_mode));
10845 + BUG_ON(bindex < dbstart(dentry));
10846 + BUG_ON(bindex > dbend(dentry));
10847 + err = is_robranch_super(sb, bindex);
10848 + if (err)
10849 + goto out;
10850 +
10851 + lower_dir_dentry = unionfs_lower_dentry_idx(dentry, bindex);
10852 + BUG_ON(!S_ISDIR(lower_dir_dentry->d_inode->i_mode));
10853 + lower_dir = lower_dir_dentry->d_inode;
10854 + BUG_ON(!S_ISDIR(lower_dir->i_mode));
10855 +
10856 + if (!inode_permission(lower_dir, MAY_WRITE | MAY_EXEC)) {
10857 + err = do_delete_whiteouts(dentry, bindex, namelist);
10858 + } else {
10859 + args.deletewh.namelist = namelist;
10860 + args.deletewh.dentry = dentry;
10861 + args.deletewh.bindex = bindex;
10862 + run_sioq(__delete_whiteouts, &args);
10863 + err = args.err;
10864 + }
10865 +
10866 +out:
10867 + return err;
10868 +}
10869 +
10870 +/****************************************************************************
10871 + * Opaque directory helpers *
10872 + ****************************************************************************/
10873 +
10874 +/*
10875 + * is_opaque_dir: returns 0 if it is NOT an opaque dir, 1 if it is, and
10876 + * -errno if an error occurred trying to figure this out.
10877 + */
10878 +int is_opaque_dir(struct dentry *dentry, int bindex)
10879 +{
10880 + int err = 0;
10881 + struct dentry *lower_dentry;
10882 + struct dentry *wh_lower_dentry;
10883 + struct inode *lower_inode;
10884 + struct sioq_args args;
10885 +
10886 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
10887 + lower_inode = lower_dentry->d_inode;
10888 +
10889 + BUG_ON(!S_ISDIR(lower_inode->i_mode));
10890 +
10891 + mutex_lock(&lower_inode->i_mutex);
10892 +
10893 + if (!inode_permission(lower_inode, MAY_EXEC)) {
10894 + wh_lower_dentry =
10895 + lookup_one_len(UNIONFS_DIR_OPAQUE, lower_dentry,
10896 + sizeof(UNIONFS_DIR_OPAQUE) - 1);
10897 + } else {
10898 + args.is_opaque.dentry = lower_dentry;
10899 + run_sioq(__is_opaque_dir, &args);
10900 + wh_lower_dentry = args.ret;
10901 + }
10902 +
10903 + mutex_unlock(&lower_inode->i_mutex);
10904 +
10905 + if (IS_ERR(wh_lower_dentry)) {
10906 + err = PTR_ERR(wh_lower_dentry);
10907 + goto out;
10908 + }
10909 +
10910 + /* This is an opaque dir iff wh_lower_dentry is positive */
10911 + err = !!wh_lower_dentry->d_inode;
10912 +
10913 + dput(wh_lower_dentry);
10914 +out:
10915 + return err;
10916 +}
10917 +
10918 +void __is_opaque_dir(struct work_struct *work)
10919 +{
10920 + struct sioq_args *args = container_of(work, struct sioq_args, work);
10921 +
10922 + args->ret = lookup_one_len(UNIONFS_DIR_OPAQUE, args->is_opaque.dentry,
10923 + sizeof(UNIONFS_DIR_OPAQUE) - 1);
10924 + complete(&args->comp);
10925 +}
10926 +
10927 +int make_dir_opaque(struct dentry *dentry, int bindex)
10928 +{
10929 + int err = 0;
10930 + struct dentry *lower_dentry, *diropq;
10931 + struct inode *lower_dir;
10932 + struct nameidata nd;
10933 + const struct cred *old_creds;
10934 + struct cred *new_creds;
10935 +
10936 + /*
10937 + * Opaque directory whiteout markers are special files (like regular
10938 + * whiteouts), and should appear to the users as if they don't
10939 + * exist. They should be created/deleted regardless of directory
10940 + * search/create permissions, but only for the duration of this
10941 + * creation of the .wh.__dir_opaque: file. Note, this does not
10942 + * circumvent normal ->permission).
10943 + */
10944 + new_creds = prepare_creds();
10945 + if (unlikely(!new_creds)) {
10946 + err = -ENOMEM;
10947 + goto out_err;
10948 + }
10949 + cap_raise(new_creds->cap_effective, CAP_DAC_READ_SEARCH);
10950 + cap_raise(new_creds->cap_effective, CAP_DAC_OVERRIDE);
10951 + old_creds = override_creds(new_creds);
10952 +
10953 + lower_dentry = unionfs_lower_dentry_idx(dentry, bindex);
10954 + lower_dir = lower_dentry->d_inode;
10955 + BUG_ON(!S_ISDIR(dentry->d_inode->i_mode) ||
10956 + !S_ISDIR(lower_dir->i_mode));
10957 +
10958 + mutex_lock(&lower_dir->i_mutex);
10959 + diropq = lookup_one_len(UNIONFS_DIR_OPAQUE, lower_dentry,
10960 + sizeof(UNIONFS_DIR_OPAQUE) - 1);
10961 + if (IS_ERR(diropq)) {
10962 + err = PTR_ERR(diropq);
10963 + goto out;
10964 + }
10965 +
10966 + err = init_lower_nd(&nd, LOOKUP_CREATE);
10967 + if (unlikely(err < 0))
10968 + goto out;
10969 + if (!diropq->d_inode)
10970 + err = vfs_create(lower_dir, diropq, S_IRUGO, &nd);
10971 + if (!err)
10972 + dbopaque(dentry) = bindex;
10973 + release_lower_nd(&nd, err);
10974 +
10975 + dput(diropq);
10976 +
10977 +out:
10978 + mutex_unlock(&lower_dir->i_mutex);
10979 + revert_creds(old_creds);
10980 +out_err:
10981 + return err;
10982 +}
10983 diff -Naur linux-2.6.30/fs/unionfs/xattr.c linux-2.6.30-magellan/fs/unionfs/xattr.c
10984 --- linux-2.6.30/fs/unionfs/xattr.c 1970-01-01 01:00:00.000000000 +0100
10985 +++ linux-2.6.30-magellan/fs/unionfs/xattr.c 2009-06-11 18:39:19.000000000 +0200
10986 @@ -0,0 +1,173 @@
10987 +/*
10988 + * Copyright (c) 2003-2009 Erez Zadok
10989 + * Copyright (c) 2003-2006 Charles P. Wright
10990 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
10991 + * Copyright (c) 2005-2006 Junjiro Okajima
10992 + * Copyright (c) 2005 Arun M. Krishnakumar
10993 + * Copyright (c) 2004-2006 David P. Quigley
10994 + * Copyright (c) 2003-2004 Mohammad Nayyer Zubair
10995 + * Copyright (c) 2003 Puja Gupta
10996 + * Copyright (c) 2003 Harikesavan Krishnan
10997 + * Copyright (c) 2003-2009 Stony Brook University
10998 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
10999 + *
11000 + * This program is free software; you can redistribute it and/or modify
11001 + * it under the terms of the GNU General Public License version 2 as
11002 + * published by the Free Software Foundation.
11003 + */
11004 +
11005 +#include "union.h"
11006 +
11007 +/* This is lifted from fs/xattr.c */
11008 +void *unionfs_xattr_alloc(size_t size, size_t limit)
11009 +{
11010 + void *ptr;
11011 +
11012 + if (size > limit)
11013 + return ERR_PTR(-E2BIG);
11014 +
11015 + if (!size) /* size request, no buffer is needed */
11016 + return NULL;
11017 +
11018 + ptr = kmalloc(size, GFP_KERNEL);
11019 + if (unlikely(!ptr))
11020 + return ERR_PTR(-ENOMEM);
11021 + return ptr;
11022 +}
11023 +
11024 +/*
11025 + * BKL held by caller.
11026 + * dentry->d_inode->i_mutex locked
11027 + */
11028 +ssize_t unionfs_getxattr(struct dentry *dentry, const char *name, void *value,
11029 + size_t size)
11030 +{
11031 + struct dentry *lower_dentry = NULL;
11032 + struct dentry *parent;
11033 + int err = -EOPNOTSUPP;
11034 + bool valid;
11035 +
11036 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
11037 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
11038 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
11039 +
11040 + valid = __unionfs_d_revalidate(dentry, parent, false);
11041 + if (unlikely(!valid)) {
11042 + err = -ESTALE;
11043 + goto out;
11044 + }
11045 +
11046 + lower_dentry = unionfs_lower_dentry(dentry);
11047 +
11048 + err = vfs_getxattr(lower_dentry, (char *) name, value, size);
11049 +
11050 +out:
11051 + unionfs_check_dentry(dentry);
11052 + unionfs_unlock_dentry(dentry);
11053 + unionfs_unlock_parent(dentry, parent);
11054 + unionfs_read_unlock(dentry->d_sb);
11055 + return err;
11056 +}
11057 +
11058 +/*
11059 + * BKL held by caller.
11060 + * dentry->d_inode->i_mutex locked
11061 + */
11062 +int unionfs_setxattr(struct dentry *dentry, const char *name,
11063 + const void *value, size_t size, int flags)
11064 +{
11065 + struct dentry *lower_dentry = NULL;
11066 + struct dentry *parent;
11067 + int err = -EOPNOTSUPP;
11068 + bool valid;
11069 +
11070 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
11071 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
11072 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
11073 +
11074 + valid = __unionfs_d_revalidate(dentry, parent, false);
11075 + if (unlikely(!valid)) {
11076 + err = -ESTALE;
11077 + goto out;
11078 + }
11079 +
11080 + lower_dentry = unionfs_lower_dentry(dentry);
11081 +
11082 + err = vfs_setxattr(lower_dentry, (char *) name, (void *) value,
11083 + size, flags);
11084 +
11085 +out:
11086 + unionfs_check_dentry(dentry);
11087 + unionfs_unlock_dentry(dentry);
11088 + unionfs_unlock_parent(dentry, parent);
11089 + unionfs_read_unlock(dentry->d_sb);
11090 + return err;
11091 +}
11092 +
11093 +/*
11094 + * BKL held by caller.
11095 + * dentry->d_inode->i_mutex locked
11096 + */
11097 +int unionfs_removexattr(struct dentry *dentry, const char *name)
11098 +{
11099 + struct dentry *lower_dentry = NULL;
11100 + struct dentry *parent;
11101 + int err = -EOPNOTSUPP;
11102 + bool valid;
11103 +
11104 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
11105 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
11106 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
11107 +
11108 + valid = __unionfs_d_revalidate(dentry, parent, false);
11109 + if (unlikely(!valid)) {
11110 + err = -ESTALE;
11111 + goto out;
11112 + }
11113 +
11114 + lower_dentry = unionfs_lower_dentry(dentry);
11115 +
11116 + err = vfs_removexattr(lower_dentry, (char *) name);
11117 +
11118 +out:
11119 + unionfs_check_dentry(dentry);
11120 + unionfs_unlock_dentry(dentry);
11121 + unionfs_unlock_parent(dentry, parent);
11122 + unionfs_read_unlock(dentry->d_sb);
11123 + return err;
11124 +}
11125 +
11126 +/*
11127 + * BKL held by caller.
11128 + * dentry->d_inode->i_mutex locked
11129 + */
11130 +ssize_t unionfs_listxattr(struct dentry *dentry, char *list, size_t size)
11131 +{
11132 + struct dentry *lower_dentry = NULL;
11133 + struct dentry *parent;
11134 + int err = -EOPNOTSUPP;
11135 + char *encoded_list = NULL;
11136 + bool valid;
11137 +
11138 + unionfs_read_lock(dentry->d_sb, UNIONFS_SMUTEX_CHILD);
11139 + parent = unionfs_lock_parent(dentry, UNIONFS_DMUTEX_PARENT);
11140 + unionfs_lock_dentry(dentry, UNIONFS_DMUTEX_CHILD);
11141 +
11142 + valid = __unionfs_d_revalidate(dentry, parent, false);
11143 + if (unlikely(!valid)) {
11144 + err = -ESTALE;
11145 + goto out;
11146 + }
11147 +
11148 + lower_dentry = unionfs_lower_dentry(dentry);
11149 +
11150 + encoded_list = list;
11151 + err = vfs_listxattr(lower_dentry, encoded_list, size);
11152 +
11153 +out:
11154 + unionfs_check_dentry(dentry);
11155 + unionfs_unlock_dentry(dentry);
11156 + unionfs_unlock_parent(dentry, parent);
11157 + unionfs_read_unlock(dentry->d_sb);
11158 + return err;
11159 +}
11160 diff -Naur linux-2.6.30/include/linux/fs_stack.h linux-2.6.30-magellan/include/linux/fs_stack.h
11161 --- linux-2.6.30/include/linux/fs_stack.h 2009-06-10 05:05:27.000000000 +0200
11162 +++ linux-2.6.30-magellan/include/linux/fs_stack.h 2009-06-11 18:39:19.000000000 +0200
11163 @@ -1,17 +1,27 @@
11164 +/*
11165 + * Copyright (c) 2006-2009 Erez Zadok
11166 + * Copyright (c) 2006-2007 Josef 'Jeff' Sipek
11167 + * Copyright (c) 2006-2009 Stony Brook University
11168 + * Copyright (c) 2006-2009 The Research Foundation of SUNY
11169 + *
11170 + * This program is free software; you can redistribute it and/or modify
11171 + * it under the terms of the GNU General Public License version 2 as
11172 + * published by the Free Software Foundation.
11173 + */
11174 +
11175 #ifndef _LINUX_FS_STACK_H
11176 #define _LINUX_FS_STACK_H
11177
11178 -/* This file defines generic functions used primarily by stackable
11179 +/*
11180 + * This file defines generic functions used primarily by stackable
11181 * filesystems; none of these functions require i_mutex to be held.
11182 */
11183
11184 #include <linux/fs.h>
11185
11186 /* externs for fs/stack.c */
11187 -extern void fsstack_copy_attr_all(struct inode *dest, const struct inode *src,
11188 - int (*get_nlinks)(struct inode *));
11189 -
11190 -extern void fsstack_copy_inode_size(struct inode *dst, const struct inode *src);
11191 +extern void fsstack_copy_attr_all(struct inode *dest, const struct inode *src);
11192 +extern void fsstack_copy_inode_size(struct inode *dst, struct inode *src);
11193
11194 /* inlines */
11195 static inline void fsstack_copy_attr_atime(struct inode *dest,
11196 diff -Naur linux-2.6.30/include/linux/magic.h linux-2.6.30-magellan/include/linux/magic.h
11197 --- linux-2.6.30/include/linux/magic.h 2009-06-10 05:05:27.000000000 +0200
11198 +++ linux-2.6.30-magellan/include/linux/magic.h 2009-06-11 18:39:19.000000000 +0200
11199 @@ -42,6 +42,8 @@
11200 #define REISER2FS_SUPER_MAGIC_STRING "ReIsEr2Fs"
11201 #define REISER2FS_JR_SUPER_MAGIC_STRING "ReIsEr3Fs"
11202
11203 +#define UNIONFS_SUPER_MAGIC 0xf15f083d
11204 +
11205 #define SMB_SUPER_MAGIC 0x517B
11206 #define USBDEVICE_SUPER_MAGIC 0x9fa2
11207 #define CGROUP_SUPER_MAGIC 0x27e0eb
11208 diff -Naur linux-2.6.30/include/linux/splice.h linux-2.6.30-magellan/include/linux/splice.h
11209 --- linux-2.6.30/include/linux/splice.h 2009-06-10 05:05:27.000000000 +0200
11210 +++ linux-2.6.30-magellan/include/linux/splice.h 2009-06-11 18:39:19.000000000 +0200
11211 @@ -82,5 +82,10 @@
11212 struct splice_pipe_desc *);
11213 extern ssize_t splice_direct_to_actor(struct file *, struct splice_desc *,
11214 splice_direct_actor *);
11215 +extern long vfs_splice_from(struct pipe_inode_info *pipe, struct file *out,
11216 + loff_t *ppos, size_t len, unsigned int flags);
11217 +extern long vfs_splice_to(struct file *in, loff_t *ppos,
11218 + struct pipe_inode_info *pipe, size_t len,
11219 + unsigned int flags);
11220
11221 #endif
11222 diff -Naur linux-2.6.30/include/linux/union_fs.h linux-2.6.30-magellan/include/linux/union_fs.h
11223 --- linux-2.6.30/include/linux/union_fs.h 1970-01-01 01:00:00.000000000 +0100
11224 +++ linux-2.6.30-magellan/include/linux/union_fs.h 2009-06-11 18:39:19.000000000 +0200
11225 @@ -0,0 +1,22 @@
11226 +/*
11227 + * Copyright (c) 2003-2009 Erez Zadok
11228 + * Copyright (c) 2005-2007 Josef 'Jeff' Sipek
11229 + * Copyright (c) 2003-2009 Stony Brook University
11230 + * Copyright (c) 2003-2009 The Research Foundation of SUNY
11231 + *
11232 + * This program is free software; you can redistribute it and/or modify
11233 + * it under the terms of the GNU General Public License version 2 as
11234 + * published by the Free Software Foundation.
11235 + */
11236 +
11237 +#ifndef _LINUX_UNION_FS_H
11238 +#define _LINUX_UNION_FS_H
11239 +
11240 +/*
11241 + * DEFINITIONS FOR USER AND KERNEL CODE:
11242 + */
11243 +# define UNIONFS_IOCTL_INCGEN _IOR(0x15, 11, int)
11244 +# define UNIONFS_IOCTL_QUERYFILE _IOR(0x15, 15, int)
11245 +
11246 +#endif /* _LINUX_UNIONFS_H */
11247 +
11248 diff -Naur linux-2.6.30/MAINTAINERS linux-2.6.30-magellan/MAINTAINERS
11249 --- linux-2.6.30/MAINTAINERS 2009-06-10 05:05:27.000000000 +0200
11250 +++ linux-2.6.30-magellan/MAINTAINERS 2009-06-11 18:39:19.000000000 +0200
11251 @@ -5723,6 +5723,14 @@
11252 F: drivers/cdrom/cdrom.c
11253 F: include/linux/cdrom.h
11254
11255 +UNIONFS
11256 +P: Erez Zadok
11257 +M: ezk@cs.sunysb.edu
11258 +L: unionfs@filesystems.org
11259 +W: http://unionfs.filesystems.org/
11260 +T: git git.kernel.org/pub/scm/linux/kernel/git/ezk/unionfs.git
11261 +S: Maintained
11262 +
11263 UNSORTED BLOCK IMAGES (UBI)
11264 P: Artem Bityutskiy
11265 M: dedekind@infradead.org
11266 diff -Naur linux-2.6.30/security/security.c linux-2.6.30-magellan/security/security.c
11267 --- linux-2.6.30/security/security.c 2009-06-10 05:05:27.000000000 +0200
11268 +++ linux-2.6.30-magellan/security/security.c 2009-06-11 18:39:19.000000000 +0200
11269 @@ -522,6 +522,7 @@
11270 return 0;
11271 return security_ops->inode_permission(inode, mask);
11272 }
11273 +EXPORT_SYMBOL(security_inode_permission);
11274
11275 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
11276 {