| 1 | |
| 2 | * Flexible array managed in PAGE_SIZE parts
|
| 3 | *
|
| 4 | * This program is free software; you can redistribute it and/or modify
|
| 5 | * it under the terms of the GNU General Public License as published by
|
| 6 | * the Free Software Foundation; either version 2 of the License, or
|
| 7 | * (at your option) any later version.
|
| 8 | *
|
| 9 | * This program is distributed in the hope that it will be useful,
|
| 10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
| 12 | * GNU General Public License for more details.
|
| 13 | *
|
| 14 | * You should have received a copy of the GNU General Public License
|
| 15 | * along with this program; if not, write to the Free Software
|
| 16 | * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
|
| 17 | *
|
| 18 | * Copyright IBM Corporation, 2009
|
| 19 | *
|
| 20 | * Author: Dave Hansen <[email protected]>
|
| 21 |
|
| 22 |
|
| 23 | #include <linux/flex_array.h>
|
| 24 | #include <linux/slab.h>
|
| 25 | #include <linux/stddef.h>
|
| 26 |
|
| 27 | struct flex_array_part {
|
| 28 | char elements[FLEX_ARRAY_PART_SIZE];
|
| 29 | };
|
| 30 |
|
| 31 | static inline int __elements_per_part(int element_size)
|
| 32 | {
|
| 33 | return FLEX_ARRAY_PART_SIZE / element_size;
|
| 34 | }
|
| 35 |
|
| 36 | static inline int bytes_left_in_base(void)
|
| 37 | {
|
| 38 | int element_offset = offsetof(struct flex_array, parts);
|
| 39 | int bytes_left = FLEX_ARRAY_BASE_SIZE - element_offset;
|
| 40 | return bytes_left;
|
| 41 | }
|
| 42 |
|
| 43 | static inline int nr_base_part_ptrs(void)
|
| 44 | {
|
| 45 | return bytes_left_in_base() / sizeof(struct flex_array_part *);
|
| 46 | }
|
| 47 |
|
| 48 | |
| 49 | * If a user requests an allocation which is small
|
| 50 | * enough, we may simply use the space in the
|
| 51 | * flex_array->parts[] array to store the user
|
| 52 | * data.
|
| 53 |
|
| 54 | static inline int elements_fit_in_base(struct flex_array *fa)
|
| 55 | {
|
| 56 | int data_size = fa->element_size * fa->total_nr_elements;
|
| 57 | if (data_size <= bytes_left_in_base())
|
| 58 | return 1;
|
| 59 | return 0;
|
| 60 | }
|
| 61 |
|
| 62 | |
| 63 | * flex_array_alloc - allocate a new flexible array
|
| 64 | * @element_size: the size of individual elements in the array
|
| 65 | * @total: total number of elements that this should hold
|
| 66 | *
|
| 67 | * Note: all locking must be provided by the caller.
|
| 68 | *
|
| 69 | * @total is used to size internal structures. If the user ever
|
| 70 | * accesses any array indexes >=@total, it will produce errors.
|
| 71 | *
|
| 72 | * The maximum number of elements is defined as: the number of
|
| 73 | * elements that can be stored in a page times the number of
|
| 74 | * page pointers that we can fit in the base structure or (using
|
| 75 | * integer math):
|
| 76 | *
|
| 77 | * (PAGE_SIZE/element_size) * (PAGE_SIZE-8)/sizeof(void *)
|
| 78 | *
|
| 79 | * Here's a table showing example capacities. Note that the maximum
|
| 80 | * index that the get/put() functions is just nr_objects-1. This
|
| 81 | * basically means that you get 4MB of storage on 32-bit and 2MB on
|
| 82 | * 64-bit.
|
| 83 | *
|
| 84 | *
|
| 85 | * Element size | Objects | Objects |
|
| 86 | * PAGE_SIZE=4k | 32-bit | 64-bit |
|
| 87 | * ---------------------------------|
|
| 88 | * 1 bytes | 4186112 | 2093056 |
|
| 89 | * 2 bytes | 2093056 | 1046528 |
|
| 90 | * 3 bytes | 1395030 | 697515 |
|
| 91 | * 4 bytes | 1046528 | 523264 |
|
| 92 | * 32 bytes | 130816 | 65408 |
|
| 93 | * 33 bytes | 126728 | 63364 |
|
| 94 | * 2048 bytes | 2044 | 1022 |
|
| 95 | * 2049 bytes | 1022 | 511 |
|
| 96 | * void * | 1046528 | 261632 |
|
| 97 | *
|
| 98 | * Since 64-bit pointers are twice the size, we lose half the
|
| 99 | * capacity in the base structure. Also note that no effort is made
|
| 100 | * to efficiently pack objects across page boundaries.
|
| 101 |
|
| 102 | struct flex_array *flex_array_alloc(int element_size, unsigned int total,
|
| 103 | gfp_t flags)
|
| 104 | {
|
| 105 | struct flex_array *ret;
|
| 106 | int max_size = nr_base_part_ptrs() * __elements_per_part(element_size);
|
| 107 |
|
| 108 |
|
| 109 | if (total > max_size)
|
| 110 | return NULL;
|
| 111 | ret = kzalloc(sizeof(struct flex_array), flags);
|
| 112 | if (!ret)
|
| 113 | return NULL;
|
| 114 | ret->element_size = element_size;
|
| 115 | ret->total_nr_elements = total;
|
| 116 | return ret;
|
| 117 | }
|
| 118 |
|
| 119 | static int fa_element_to_part_nr(struct flex_array *fa,
|
| 120 | unsigned int element_nr)
|
| 121 | {
|
| 122 | return element_nr / __elements_per_part(fa->element_size);
|
| 123 | }
|
| 124 |
|
| 125 | |
| 126 | * flex_array_free_parts - just free the second-level pages
|
| 127 | *
|
| 128 | * This is to be used in cases where the base 'struct flex_array'
|
| 129 | * has been statically allocated and should not be free.
|
| 130 |
|
| 131 | void flex_array_free_parts(struct flex_array *fa)
|
| 132 | {
|
| 133 | int part_nr;
|
| 134 | int max_part = nr_base_part_ptrs();
|
| 135 |
|
| 136 | if (elements_fit_in_base(fa))
|
| 137 | return;
|
| 138 | for (part_nr = 0; part_nr < max_part; part_nr++)
|
| 139 | kfree(fa->parts[part_nr]);
|
| 140 | }
|
| 141 |
|
| 142 | void flex_array_free(struct flex_array *fa)
|
| 143 | {
|
| 144 | flex_array_free_parts(fa);
|
| 145 | kfree(fa);
|
| 146 | }
|
| 147 |
|
| 148 | static unsigned int index_inside_part(struct flex_array *fa,
|
| 149 | unsigned int element_nr)
|
| 150 | {
|
| 151 | unsigned int part_offset;
|
| 152 |
|
| 153 | part_offset = element_nr % __elements_per_part(fa->element_size);
|
| 154 | return part_offset * fa->element_size;
|
| 155 | }
|
| 156 |
|
| 157 | static struct flex_array_part *
|
| 158 | __fa_get_part(struct flex_array *fa, int part_nr, gfp_t flags)
|
| 159 | {
|
| 160 | struct flex_array_part *part = fa->parts[part_nr];
|
| 161 | if (!part) {
|
| 162 | |
| 163 | * This leaves the part pages uninitialized
|
| 164 | * and with potentially random data, just
|
| 165 | * as if the user had kmalloc()'d the whole.
|
| 166 | * __GFP_ZERO can be used to zero it.
|
| 167 |
|
| 168 | part = kmalloc(FLEX_ARRAY_PART_SIZE, flags);
|
| 169 | if (!part)
|
| 170 | return NULL;
|
| 171 | fa->parts[part_nr] = part;
|
| 172 | }
|
| 173 | return part;
|
| 174 | }
|
| 175 |
|
| 176 | |
| 177 | * flex_array_put - copy data into the array at @element_nr
|
| 178 | * @src: address of data to copy into the array
|
| 179 | * @element_nr: index of the position in which to insert
|
| 180 | * the new element.
|
| 181 | *
|
| 182 | * Note that this *copies* the contents of @src into
|
| 183 | * the array. If you are trying to store an array of
|
| 184 | * pointers, make sure to pass in &ptr instead of ptr.
|
| 185 | *
|
| 186 | * Locking must be provided by the caller.
|
| 187 |
|
| 188 | int flex_array_put(struct flex_array *fa, unsigned int element_nr, void *src,
|
| 189 | gfp_t flags)
|
| 190 | {
|
| 191 | int part_nr = fa_element_to_part_nr(fa, element_nr);
|
| 192 | struct flex_array_part *part;
|
| 193 | void *dst;
|
| 194 |
|
| 195 | if (element_nr >= fa->total_nr_elements)
|
| 196 | return -ENOSPC;
|
| 197 | if (elements_fit_in_base(fa))
|
| 198 | part = (struct flex_array_part *)&fa->parts[0];
|
| 199 | else {
|
| 200 | part = __fa_get_part(fa, part_nr, flags);
|
| 201 | if (!part)
|
| 202 | return -ENOMEM;
|
| 203 | }
|
| 204 | dst = &part->elements[index_inside_part(fa, element_nr)];
|
| 205 | memcpy(dst, src, fa->element_size);
|
| 206 | return 0;
|
| 207 | }
|
| 208 |
|
| 209 | |
| 210 | * flex_array_prealloc - guarantee that array space exists
|
| 211 | * @start: index of first array element for which space is allocated
|
| 212 | * @end: index of last (inclusive) element for which space is allocated
|
| 213 | *
|
| 214 | * This will guarantee that no future calls to flex_array_put()
|
| 215 | * will allocate memory. It can be used if you are expecting to
|
| 216 | * be holding a lock or in some atomic context while writing
|
| 217 | * data into the array.
|
| 218 | *
|
| 219 | * Locking must be provided by the caller.
|
| 220 |
|
| 221 | int flex_array_prealloc(struct flex_array *fa, unsigned int start,
|
| 222 | unsigned int end, gfp_t flags)
|
| 223 | {
|
| 224 | int start_part;
|
| 225 | int end_part;
|
| 226 | int part_nr;
|
| 227 | struct flex_array_part *part;
|
| 228 |
|
| 229 | if (start >= fa->total_nr_elements || end >= fa->total_nr_elements)
|
| 230 | return -ENOSPC;
|
| 231 | if (elements_fit_in_base(fa))
|
| 232 | return 0;
|
| 233 | start_part = fa_element_to_part_nr(fa, start);
|
| 234 | end_part = fa_element_to_part_nr(fa, end);
|
| 235 | for (part_nr = start_part; part_nr <= end_part; part_nr++) {
|
| 236 | part = __fa_get_part(fa, part_nr, flags);
|
| 237 | if (!part)
|
| 238 | return -ENOMEM;
|
| 239 | }
|
| 240 | return 0;
|
| 241 | }
|
| 242 |
|
| 243 | |
| 244 | * flex_array_get - pull data back out of the array
|
| 245 | * @element_nr: index of the element to fetch from the array
|
| 246 | *
|
| 247 | * Returns a pointer to the data at index @element_nr. Note
|
| 248 | * that this is a copy of the data that was passed in. If you
|
| 249 | * are using this to store pointers, you'll get back &ptr.
|
| 250 | *
|
| 251 | * Locking must be provided by the caller.
|
| 252 |
|
| 253 | void *flex_array_get(struct flex_array *fa, unsigned int element_nr)
|
| 254 | {
|
| 255 | int part_nr = fa_element_to_part_nr(fa, element_nr);
|
| 256 | struct flex_array_part *part;
|
| 257 |
|
| 258 | if (element_nr >= fa->total_nr_elements)
|
| 259 | return NULL;
|
| 260 | if (elements_fit_in_base(fa))
|
| 261 | part = (struct flex_array_part *)&fa->parts[0];
|
| 262 | else {
|
| 263 | part = fa->parts[part_nr];
|
| 264 | if (!part)
|
| 265 | return NULL;
|
| 266 | }
|
| 267 | return &part->elements[index_inside_part(fa, element_nr)];
|
| 268 | }
|