]>
Commit | Line | Data |
---|---|---|
1 | // SPDX-License-Identifier: GPL-2.0 | |
2 | /* | |
3 | * quota.c - CephFS quota | |
4 | * | |
5 | * Copyright (C) 2017-2018 SUSE | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License | |
9 | * as published by the Free Software Foundation; either version 2 | |
10 | * of the License, or (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, see <http://www.gnu.org/licenses/>. | |
19 | */ | |
20 | ||
21 | #include "super.h" | |
22 | #include "mds_client.h" | |
23 | ||
24 | static inline bool ceph_has_quota(struct ceph_inode_info *ci) | |
25 | { | |
26 | return (ci && (ci->i_max_files || ci->i_max_bytes)); | |
27 | } | |
28 | ||
29 | void ceph_handle_quota(struct ceph_mds_client *mdsc, | |
30 | struct ceph_mds_session *session, | |
31 | struct ceph_msg *msg) | |
32 | { | |
33 | struct super_block *sb = mdsc->fsc->sb; | |
34 | struct ceph_mds_quota *h = msg->front.iov_base; | |
35 | struct ceph_vino vino; | |
36 | struct inode *inode; | |
37 | struct ceph_inode_info *ci; | |
38 | ||
39 | if (msg->front.iov_len != sizeof(*h)) { | |
40 | pr_err("%s corrupt message mds%d len %d\n", __func__, | |
41 | session->s_mds, (int)msg->front.iov_len); | |
42 | ceph_msg_dump(msg); | |
43 | return; | |
44 | } | |
45 | ||
46 | /* increment msg sequence number */ | |
47 | mutex_lock(&session->s_mutex); | |
48 | session->s_seq++; | |
49 | mutex_unlock(&session->s_mutex); | |
50 | ||
51 | /* lookup inode */ | |
52 | vino.ino = le64_to_cpu(h->ino); | |
53 | vino.snap = CEPH_NOSNAP; | |
54 | inode = ceph_find_inode(sb, vino); | |
55 | if (!inode) { | |
56 | pr_warn("Failed to find inode %llu\n", vino.ino); | |
57 | return; | |
58 | } | |
59 | ci = ceph_inode(inode); | |
60 | ||
61 | spin_lock(&ci->i_ceph_lock); | |
62 | ci->i_rbytes = le64_to_cpu(h->rbytes); | |
63 | ci->i_rfiles = le64_to_cpu(h->rfiles); | |
64 | ci->i_rsubdirs = le64_to_cpu(h->rsubdirs); | |
65 | ci->i_max_bytes = le64_to_cpu(h->max_bytes); | |
66 | ci->i_max_files = le64_to_cpu(h->max_files); | |
67 | spin_unlock(&ci->i_ceph_lock); | |
68 | ||
69 | iput(inode); | |
70 | } | |
71 | ||
72 | /* | |
73 | * This function walks through the snaprealm for an inode and returns the | |
74 | * ceph_snap_realm for the first snaprealm that has quotas set (either max_files | |
75 | * or max_bytes). If the root is reached, return the root ceph_snap_realm | |
76 | * instead. | |
77 | * | |
78 | * Note that the caller is responsible for calling ceph_put_snap_realm() on the | |
79 | * returned realm. | |
80 | */ | |
81 | static struct ceph_snap_realm *get_quota_realm(struct ceph_mds_client *mdsc, | |
82 | struct inode *inode) | |
83 | { | |
84 | struct ceph_inode_info *ci = NULL; | |
85 | struct ceph_snap_realm *realm, *next; | |
86 | struct ceph_vino vino; | |
87 | struct inode *in; | |
88 | ||
89 | if (ceph_snap(inode) != CEPH_NOSNAP) | |
90 | return NULL; | |
91 | ||
92 | realm = ceph_inode(inode)->i_snap_realm; | |
93 | if (realm) | |
94 | ceph_get_snap_realm(mdsc, realm); | |
95 | else | |
96 | pr_err_ratelimited("get_quota_realm: ino (%llx.%llx) " | |
97 | "null i_snap_realm\n", ceph_vinop(inode)); | |
98 | while (realm) { | |
99 | vino.ino = realm->ino; | |
100 | vino.snap = CEPH_NOSNAP; | |
101 | in = ceph_find_inode(inode->i_sb, vino); | |
102 | if (!in) { | |
103 | pr_warn("Failed to find inode for %llu\n", vino.ino); | |
104 | break; | |
105 | } | |
106 | ci = ceph_inode(in); | |
107 | if (ceph_has_quota(ci) || (ci->i_vino.ino == CEPH_INO_ROOT)) { | |
108 | iput(in); | |
109 | return realm; | |
110 | } | |
111 | iput(in); | |
112 | next = realm->parent; | |
113 | ceph_get_snap_realm(mdsc, next); | |
114 | ceph_put_snap_realm(mdsc, realm); | |
115 | realm = next; | |
116 | } | |
117 | if (realm) | |
118 | ceph_put_snap_realm(mdsc, realm); | |
119 | ||
120 | return NULL; | |
121 | } | |
122 | ||
123 | bool ceph_quota_is_same_realm(struct inode *old, struct inode *new) | |
124 | { | |
125 | struct ceph_mds_client *mdsc = ceph_inode_to_client(old)->mdsc; | |
126 | struct ceph_snap_realm *old_realm, *new_realm; | |
127 | bool is_same; | |
128 | ||
129 | down_read(&mdsc->snap_rwsem); | |
130 | old_realm = get_quota_realm(mdsc, old); | |
131 | new_realm = get_quota_realm(mdsc, new); | |
132 | is_same = (old_realm == new_realm); | |
133 | up_read(&mdsc->snap_rwsem); | |
134 | ||
135 | if (old_realm) | |
136 | ceph_put_snap_realm(mdsc, old_realm); | |
137 | if (new_realm) | |
138 | ceph_put_snap_realm(mdsc, new_realm); | |
139 | ||
140 | return is_same; | |
141 | } | |
142 | ||
143 | enum quota_check_op { | |
144 | QUOTA_CHECK_MAX_FILES_OP, /* check quota max_files limit */ | |
145 | QUOTA_CHECK_MAX_BYTES_OP, /* check quota max_files limit */ | |
146 | QUOTA_CHECK_MAX_BYTES_APPROACHING_OP /* check if quota max_files | |
147 | limit is approaching */ | |
148 | }; | |
149 | ||
150 | /* | |
151 | * check_quota_exceeded() will walk up the snaprealm hierarchy and, for each | |
152 | * realm, it will execute quota check operation defined by the 'op' parameter. | |
153 | * The snaprealm walk is interrupted if the quota check detects that the quota | |
154 | * is exceeded or if the root inode is reached. | |
155 | */ | |
156 | static bool check_quota_exceeded(struct inode *inode, enum quota_check_op op, | |
157 | loff_t delta) | |
158 | { | |
159 | struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc; | |
160 | struct ceph_inode_info *ci; | |
161 | struct ceph_snap_realm *realm, *next; | |
162 | struct ceph_vino vino; | |
163 | struct inode *in; | |
164 | u64 max, rvalue; | |
165 | bool is_root; | |
166 | bool exceeded = false; | |
167 | ||
168 | if (ceph_snap(inode) != CEPH_NOSNAP) | |
169 | return false; | |
170 | ||
171 | down_read(&mdsc->snap_rwsem); | |
172 | realm = ceph_inode(inode)->i_snap_realm; | |
173 | if (realm) | |
174 | ceph_get_snap_realm(mdsc, realm); | |
175 | else | |
176 | pr_err_ratelimited("check_quota_exceeded: ino (%llx.%llx) " | |
177 | "null i_snap_realm\n", ceph_vinop(inode)); | |
178 | while (realm) { | |
179 | vino.ino = realm->ino; | |
180 | vino.snap = CEPH_NOSNAP; | |
181 | in = ceph_find_inode(inode->i_sb, vino); | |
182 | if (!in) { | |
183 | pr_warn("Failed to find inode for %llu\n", vino.ino); | |
184 | break; | |
185 | } | |
186 | ci = ceph_inode(in); | |
187 | spin_lock(&ci->i_ceph_lock); | |
188 | if (op == QUOTA_CHECK_MAX_FILES_OP) { | |
189 | max = ci->i_max_files; | |
190 | rvalue = ci->i_rfiles + ci->i_rsubdirs; | |
191 | } else { | |
192 | max = ci->i_max_bytes; | |
193 | rvalue = ci->i_rbytes; | |
194 | } | |
195 | is_root = (ci->i_vino.ino == CEPH_INO_ROOT); | |
196 | spin_unlock(&ci->i_ceph_lock); | |
197 | switch (op) { | |
198 | case QUOTA_CHECK_MAX_FILES_OP: | |
199 | exceeded = (max && (rvalue >= max)); | |
200 | break; | |
201 | case QUOTA_CHECK_MAX_BYTES_OP: | |
202 | exceeded = (max && (rvalue + delta > max)); | |
203 | break; | |
204 | case QUOTA_CHECK_MAX_BYTES_APPROACHING_OP: | |
205 | if (max) { | |
206 | if (rvalue >= max) | |
207 | exceeded = true; | |
208 | else { | |
209 | /* | |
210 | * when we're writing more that 1/16th | |
211 | * of the available space | |
212 | */ | |
213 | exceeded = | |
214 | (((max - rvalue) >> 4) < delta); | |
215 | } | |
216 | } | |
217 | break; | |
218 | default: | |
219 | /* Shouldn't happen */ | |
220 | pr_warn("Invalid quota check op (%d)\n", op); | |
221 | exceeded = true; /* Just break the loop */ | |
222 | } | |
223 | iput(in); | |
224 | ||
225 | if (is_root || exceeded) | |
226 | break; | |
227 | next = realm->parent; | |
228 | ceph_get_snap_realm(mdsc, next); | |
229 | ceph_put_snap_realm(mdsc, realm); | |
230 | realm = next; | |
231 | } | |
232 | ceph_put_snap_realm(mdsc, realm); | |
233 | up_read(&mdsc->snap_rwsem); | |
234 | ||
235 | return exceeded; | |
236 | } | |
237 | ||
238 | /* | |
239 | * ceph_quota_is_max_files_exceeded - check if we can create a new file | |
240 | * @inode: directory where a new file is being created | |
241 | * | |
242 | * This functions returns true is max_files quota allows a new file to be | |
243 | * created. It is necessary to walk through the snaprealm hierarchy (until the | |
244 | * FS root) to check all realms with quotas set. | |
245 | */ | |
246 | bool ceph_quota_is_max_files_exceeded(struct inode *inode) | |
247 | { | |
248 | WARN_ON(!S_ISDIR(inode->i_mode)); | |
249 | ||
250 | return check_quota_exceeded(inode, QUOTA_CHECK_MAX_FILES_OP, 0); | |
251 | } | |
252 | ||
253 | /* | |
254 | * ceph_quota_is_max_bytes_exceeded - check if we can write to a file | |
255 | * @inode: inode being written | |
256 | * @newsize: new size if write succeeds | |
257 | * | |
258 | * This functions returns true is max_bytes quota allows a file size to reach | |
259 | * @newsize; it returns false otherwise. | |
260 | */ | |
261 | bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, loff_t newsize) | |
262 | { | |
263 | loff_t size = i_size_read(inode); | |
264 | ||
265 | /* return immediately if we're decreasing file size */ | |
266 | if (newsize <= size) | |
267 | return false; | |
268 | ||
269 | return check_quota_exceeded(inode, QUOTA_CHECK_MAX_BYTES_OP, (newsize - size)); | |
270 | } | |
271 | ||
272 | /* | |
273 | * ceph_quota_is_max_bytes_approaching - check if we're reaching max_bytes | |
274 | * @inode: inode being written | |
275 | * @newsize: new size if write succeeds | |
276 | * | |
277 | * This function returns true if the new file size @newsize will be consuming | |
278 | * more than 1/16th of the available quota space; it returns false otherwise. | |
279 | */ | |
280 | bool ceph_quota_is_max_bytes_approaching(struct inode *inode, loff_t newsize) | |
281 | { | |
282 | loff_t size = ceph_inode(inode)->i_reported_size; | |
283 | ||
284 | /* return immediately if we're decreasing file size */ | |
285 | if (newsize <= size) | |
286 | return false; | |
287 | ||
288 | return check_quota_exceeded(inode, QUOTA_CHECK_MAX_BYTES_APPROACHING_OP, | |
289 | (newsize - size)); | |
290 | } |