7754 need tmpfs size support in percent
[unleashed.git] / usr / src / uts / common / fs / tmpfs / tmp_subr.c
blob87236315554f6ba16fcfff42bf2c126b79866cf3
1 /*
2 * CDDL HEADER START
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
19 * CDDL HEADER END
22 * Copyright (c) 1990, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2015 Joyent, Inc.
26 #include <sys/types.h>
27 #include <sys/errno.h>
28 #include <sys/param.h>
29 #include <sys/t_lock.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/debug.h>
33 #include <sys/time.h>
34 #include <sys/cmn_err.h>
35 #include <sys/vnode.h>
36 #include <sys/stat.h>
37 #include <sys/vfs.h>
38 #include <sys/cred.h>
39 #include <sys/kmem.h>
40 #include <sys/atomic.h>
41 #include <sys/policy.h>
42 #include <sys/fs/tmp.h>
43 #include <sys/fs/tmpnode.h>
44 #include <sys/ddi.h>
45 #include <sys/sunddi.h>
47 #define KILOBYTE 1024
48 #define MEGABYTE (1024 * KILOBYTE)
49 #define GIGABYTE (1024 * MEGABYTE)
51 #define MODESHIFT 3
53 #define VALIDMODEBITS 07777
55 extern pgcnt_t swapfs_minfree;
57 int
58 tmp_taccess(void *vtp, int mode, struct cred *cred)
60 struct tmpnode *tp = vtp;
61 int shift = 0;
63 * Check access based on owner, group and
64 * public permissions in tmpnode.
66 if (crgetuid(cred) != tp->tn_uid) {
67 shift += MODESHIFT;
68 if (groupmember(tp->tn_gid, cred) == 0)
69 shift += MODESHIFT;
72 return (secpolicy_vnode_access2(cred, TNTOV(tp), tp->tn_uid,
73 tp->tn_mode << shift, mode));
77 * Decide whether it is okay to remove within a sticky directory.
78 * Two conditions need to be met: write access to the directory
79 * is needed. In sticky directories, write access is not sufficient;
80 * you can remove entries from a directory only if you own the directory,
81 * if you are privileged, if you own the entry or if they entry is
82 * a plain file and you have write access to that file.
83 * Function returns 0 if remove access is granted.
85 int
86 tmp_sticky_remove_access(struct tmpnode *dir, struct tmpnode *entry,
87 struct cred *cr)
89 uid_t uid = crgetuid(cr);
91 if ((dir->tn_mode & S_ISVTX) &&
92 uid != dir->tn_uid &&
93 uid != entry->tn_uid &&
94 (entry->tn_type != VREG ||
95 tmp_taccess(entry, VWRITE, cr) != 0))
96 return (secpolicy_vnode_remove(cr));
98 return (0);
102 * Allocate zeroed memory if tmpfs_maxkmem has not been exceeded
103 * or the 'musthave' flag is set. 'musthave' allocations should
104 * always be subordinate to normal allocations so that tmpfs_maxkmem
105 * can't be exceeded by more than a few KB. Example: when creating
106 * a new directory, the tmpnode is a normal allocation; if that
107 * succeeds, the dirents for "." and ".." are 'musthave' allocations.
109 void *
110 tmp_memalloc(size_t size, int musthave)
112 static time_t last_warning;
113 time_t now;
115 if (atomic_add_long_nv(&tmp_kmemspace, size) < tmpfs_maxkmem ||
116 musthave)
117 return (kmem_zalloc(size, KM_SLEEP));
119 atomic_add_long(&tmp_kmemspace, -size);
120 now = gethrestime_sec();
121 if (last_warning != now) {
122 last_warning = now;
123 cmn_err(CE_WARN, "tmp_memalloc: tmpfs over memory limit");
125 return (NULL);
128 void
129 tmp_memfree(void *cp, size_t size)
131 kmem_free(cp, size);
132 atomic_add_long(&tmp_kmemspace, -size);
136 * Convert a string containing a number (number of bytes) to a pgcnt_t,
137 * containing the corresponding number of pages. On 32-bit kernels, the
138 * maximum value encoded in 'str' is PAGESIZE * ULONG_MAX, while the value
139 * returned in 'maxpg' is at most ULONG_MAX.
141 * The number may be followed by a magnitude suffix: "k" or "K" for kilobytes;
142 * "m" or "M" for megabytes; "g" or "G" for gigabytes. This interface allows
143 * for an arguably esoteric interpretation of multiple suffix characters:
144 * namely, they cascade. For example, the caller may specify "2mk", which is
145 * interpreted as 2 gigabytes. It would seem, at this late stage, that the
146 * horse has left not only the barn but indeed the country, and possibly the
147 * entire planetary system. Alternatively, the number may be followed by a
148 * single '%' sign, indicating the size is a percentage of either the zone's
149 * swap limit or the system's overall swap size.
151 * Parse and overflow errors are detected and a non-zero number returned on
152 * error.
155 tmp_convnum(char *str, pgcnt_t *maxpg)
157 u_longlong_t num = 0;
158 #ifdef _LP64
159 u_longlong_t max_bytes = ULONG_MAX;
160 #else
161 u_longlong_t max_bytes = PAGESIZE * (uint64_t)ULONG_MAX;
162 #endif
163 char *c;
164 const struct convchar {
165 char *cc_char;
166 uint64_t cc_factor;
167 } convchars[] = {
168 { "kK", KILOBYTE },
169 { "mM", MEGABYTE },
170 { "gG", GIGABYTE },
171 { NULL, 0 }
174 if (str == NULL) {
175 return (EINVAL);
177 c = str;
180 * Convert the initial numeric portion of the input string.
182 if (ddi_strtoull(str, &c, 10, &num) != 0) {
183 return (EINVAL);
187 * Handle a size in percent. Anything other than a single percent
188 * modifier is invalid. We use either the zone's swap limit or the
189 * system's total available swap size as the initial value. Perform the
190 * intermediate calculation in pages to avoid overflow.
192 if (*c == '%') {
193 u_longlong_t cap;
195 if (*(c + 1) != '\0')
196 return (EINVAL);
198 if (num > 100)
199 return (EINVAL);
201 cap = (u_longlong_t)curproc->p_zone->zone_max_swap_ctl;
202 if (cap == UINT64_MAX) {
204 * Use the amount of available physical and memory swap
206 mutex_enter(&anoninfo_lock);
207 cap = TOTAL_AVAILABLE_SWAP;
208 mutex_exit(&anoninfo_lock);
209 } else {
210 cap = btop(cap);
213 num = ptob(cap * num / 100);
214 goto done;
218 * Apply the (potentially cascading) magnitude suffixes until an
219 * invalid character is found, or the string comes to an end.
221 for (; *c != '\0'; c++) {
222 int i;
224 for (i = 0; convchars[i].cc_char != NULL; i++) {
226 * Check if this character matches this multiplier
227 * class:
229 if (strchr(convchars[i].cc_char, *c) != NULL) {
231 * Check for overflow:
233 if (num > max_bytes / convchars[i].cc_factor) {
234 return (EINVAL);
237 num *= convchars[i].cc_factor;
238 goto valid_char;
243 * This was not a valid multiplier suffix character.
245 return (EINVAL);
247 valid_char:
248 continue;
251 done:
253 * Since btopr() rounds up to page granularity, this round-up can
254 * cause an overflow only if 'num' is between (max_bytes - PAGESIZE)
255 * and (max_bytes). In this case the resulting number is zero, which
256 * is what we check for below.
258 if ((*maxpg = (pgcnt_t)btopr(num)) == 0 && num != 0)
259 return (EINVAL);
260 return (0);
264 * Parse an octal mode string for use as the permissions set for the root
265 * of the tmpfs mount.
268 tmp_convmode(char *str, mode_t *mode)
270 ulong_t num;
271 char *c;
273 if (str == NULL) {
274 return (EINVAL);
277 if (ddi_strtoul(str, &c, 8, &num) != 0) {
278 return (EINVAL);
281 if ((num & ~VALIDMODEBITS) != 0) {
282 return (EINVAL);
285 *mode = VALIDMODEBITS & num;
286 return (0);