aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorArtem B. Bityutskiy <dedekind@linutronix.de>2006-06-27 04:22:22 -0400
committerFrank Haverkamp <haver@vnet.ibm.com>2007-04-27 07:23:33 -0400
commit801c135ce73d5df1caf3eca35b66a10824ae0707 (patch)
treeeaf6e7859650557192533b70746479de686c56e1
parentde46c33745f5e2ad594c72f2cf5f490861b16ce1 (diff)
UBI: Unsorted Block Images
UBI (Latin: "where?") manages multiple logical volumes on a single flash device, specifically supporting NAND flash devices. UBI provides a flexible partitioning concept which still allows for wear-levelling across the whole flash device. In a sense, UBI may be compared to the Logical Volume Manager (LVM). Whereas LVM maps logical sector numbers to physical HDD sector numbers, UBI maps logical eraseblocks to physical eraseblocks. More information may be found at http://www.linux-mtd.infradead.org/doc/ubi.html Partitioning/Re-partitioning An UBI volume occupies a certain number of erase blocks. This is limited by a configured maximum volume size, which could also be viewed as the partition size. Each individual UBI volume's size can be changed independently of the other UBI volumes, provided that the sum of all volume sizes doesn't exceed a certain limit. UBI supports dynamic volumes and static volumes. Static volumes are read-only and their contents are protected by CRC check sums. Bad eraseblocks handling UBI transparently handles bad eraseblocks. When a physical eraseblock becomes bad, it is substituted by a good physical eraseblock, and the user does not even notice this. Scrubbing On a NAND flash bit flips can occur on any write operation, sometimes also on read. If bit flips persist on the device, at first they can still be corrected by ECC, but once they accumulate, correction will become impossible. Thus it is best to actively scrub the affected eraseblock, by first copying it to a free eraseblock and then erasing the original. The UBI layer performs this type of scrubbing under the covers, transparently to the UBI volume users. Erase Counts UBI maintains an erase count header per eraseblock. This frees higher-level layers (like file systems) from doing this and allows for centralized erase count management instead. The erase counts are used by the wear-levelling algorithm in the UBI layer. The algorithm itself is exchangeable. Booting from NAND For booting directly from NAND flash the hardware must at least be capable of fetching and executing a small portion of the NAND flash. Some NAND flash controllers have this kind of support. They usually limit the window to a few kilobytes in erase block 0. This "initial program loader" (IPL) must then contain sufficient logic to load and execute the next boot phase. Due to bad eraseblocks, which may be randomly scattered over the flash device, it is problematic to store the "secondary program loader" (SPL) statically. Also, due to bit-flips it may become corrupted over time. UBI allows to solve this problem gracefully by storing the SPL in a small static UBI volume. UBI volumes vs. static partitions UBI volumes are still very similar to static MTD partitions: * both consist of eraseblocks (logical eraseblocks in case of UBI volumes, and physical eraseblocks in case of static partitions; * both support three basic operations - read, write, erase. But UBI volumes have the following advantages over traditional static MTD partitions: * there are no eraseblock wear-leveling constraints in case of UBI volumes, so the user should not care about this; * there are no bit-flips and bad eraseblocks in case of UBI volumes. So, UBI volumes may be considered as flash devices with relaxed restrictions. Where can it be found? Documentation, kernel code and applications can be found in the MTD gits. What are the applications for? The applications help to create binary flash images for two purposes: pfi files (partial flash images) for in-system update of UBI volumes, and plain binary images, with or without OOB data in case of NAND, for a manufacturing step. Furthermore some tools are/and will be created that allow flash content analysis after a system has crashed.. Who did UBI? The original ideas, where UBI is based on, were developed by Andreas Arnez, Frank Haverkamp and Thomas Gleixner. Josh W. Boyer and some others were involved too. The implementation of the kernel layer was done by Artem B. Bityutskiy. The user-space applications and tools were written by Oliver Lohmann with contributions from Frank Haverkamp, Andreas Arnez, and Artem. Joern Engel contributed a patch which modifies JFFS2 so that it can be run on a UBI volume. Thomas Gleixner did modifications to the NAND layer. Alexander Schmidt made some testing work as well as core functionality improvements. Signed-off-by: Artem B. Bityutskiy <dedekind@linutronix.de> Signed-off-by: Frank Haverkamp <haver@vnet.ibm.com>
-rw-r--r--drivers/mtd/Kconfig2
-rw-r--r--drivers/mtd/Makefile2
-rw-r--r--drivers/mtd/ubi/Kconfig58
-rw-r--r--drivers/mtd/ubi/Kconfig.debug104
-rw-r--r--drivers/mtd/ubi/Makefile7
-rw-r--r--drivers/mtd/ubi/build.c848
-rw-r--r--drivers/mtd/ubi/cdev.c722
-rw-r--r--drivers/mtd/ubi/debug.c224
-rw-r--r--drivers/mtd/ubi/debug.h161
-rw-r--r--drivers/mtd/ubi/eba.c1241
-rw-r--r--drivers/mtd/ubi/gluebi.c324
-rw-r--r--drivers/mtd/ubi/io.c1259
-rw-r--r--drivers/mtd/ubi/kapi.c575
-rw-r--r--drivers/mtd/ubi/misc.c105
-rw-r--r--drivers/mtd/ubi/scan.c1368
-rw-r--r--drivers/mtd/ubi/scan.h167
-rw-r--r--drivers/mtd/ubi/ubi.h535
-rw-r--r--drivers/mtd/ubi/upd.c348
-rw-r--r--drivers/mtd/ubi/vmt.c809
-rw-r--r--drivers/mtd/ubi/vtbl.c809
-rw-r--r--drivers/mtd/ubi/wl.c1671
-rw-r--r--include/linux/mtd/ubi.h202
-rw-r--r--include/mtd/Kbuild2
-rw-r--r--include/mtd/mtd-abi.h1
-rw-r--r--include/mtd/ubi-header.h360
-rw-r--r--include/mtd/ubi-user.h161
26 files changed, 12065 insertions, 0 deletions
diff --git a/drivers/mtd/Kconfig b/drivers/mtd/Kconfig
index 26f75c299440..6d1b91bf7ad5 100644
--- a/drivers/mtd/Kconfig
+++ b/drivers/mtd/Kconfig
@@ -292,5 +292,7 @@ source "drivers/mtd/nand/Kconfig"
292 292
293source "drivers/mtd/onenand/Kconfig" 293source "drivers/mtd/onenand/Kconfig"
294 294
295source "drivers/mtd/ubi/Kconfig"
296
295endmenu 297endmenu
296 298
diff --git a/drivers/mtd/Makefile b/drivers/mtd/Makefile
index c130e6261adf..92055405cb30 100644
--- a/drivers/mtd/Makefile
+++ b/drivers/mtd/Makefile
@@ -28,3 +28,5 @@ nftl-objs := nftlcore.o nftlmount.o
28inftl-objs := inftlcore.o inftlmount.o 28inftl-objs := inftlcore.o inftlmount.o
29 29
30obj-y += chips/ maps/ devices/ nand/ onenand/ 30obj-y += chips/ maps/ devices/ nand/ onenand/
31
32obj-$(CONFIG_MTD_UBI) += ubi/
diff --git a/drivers/mtd/ubi/Kconfig b/drivers/mtd/ubi/Kconfig
new file mode 100644
index 000000000000..b9daf159a4a7
--- /dev/null
+++ b/drivers/mtd/ubi/Kconfig
@@ -0,0 +1,58 @@
1# drivers/mtd/ubi/Kconfig
2
3menu "UBI - Unsorted block images"
4 depends on MTD
5
6config MTD_UBI
7 tristate "Enable UBI"
8 depends on MTD
9 select CRC32
10 help
11 UBI is a software layer above MTD layer which admits of LVM-like
12 logical volumes on top of MTD devices, hides some complexities of
13 flash chips like wear and bad blocks and provides some other useful
14 capabilities. Please, consult the MTD web site for more details
15 (www.linux-mtd.infradead.org).
16
17config MTD_UBI_WL_THRESHOLD
18 int "UBI wear-leveling threshold"
19 default 4096
20 range 2 65536
21 depends on MTD_UBI
22 help
23 This parameter defines the maximum difference between the highest
24 erase counter value and the lowest erase counter value of eraseblocks
25 of UBI devices. When this threshold is exceeded, UBI starts performing
26 wear leveling by means of moving data from eraseblock with low erase
27 counter to eraseblocks with high erase counter. Leave the default
28 value if unsure.
29
30config MTD_UBI_BEB_RESERVE
31 int "Percentage of reserved eraseblocks for bad eraseblocks handling"
32 default 1
33 range 0 25
34 depends on MTD_UBI
35 help
36 If the MTD device admits of bad eraseblocks (e.g. NAND flash), UBI
37 reserves some amount of physical eraseblocks to handle new bad
38 eraseblocks. For example, if a flash physical eraseblock becomes bad,
39 UBI uses these reserved physical eraseblocks to relocate the bad one.
40 This option specifies how many physical eraseblocks will be reserved
41 for bad eraseblock handling (percents of total number of good flash
42 eraseblocks). If the underlying flash does not admit of bad
43 eraseblocks (e.g. NOR flash), this value is ignored and nothing is
44 reserved. Leave the default value if unsure.
45
46config MTD_UBI_GLUEBI
47 bool "Emulate MTD devices"
48 default n
49 depends on MTD_UBI
50 help
51 This option enables MTD devices emulation on top of UBI volumes: for
52 each UBI volumes an MTD device is created, and all I/O to this MTD
53 device is redirected to the UBI volume. This is handy to make
54 MTD-oriented software (like JFFS2) work on top of UBI. Do not enable
55 this if no legacy software will be used.
56
57source "drivers/mtd/ubi/Kconfig.debug"
58endmenu
diff --git a/drivers/mtd/ubi/Kconfig.debug b/drivers/mtd/ubi/Kconfig.debug
new file mode 100644
index 000000000000..1e2ee22edeff
--- /dev/null
+++ b/drivers/mtd/ubi/Kconfig.debug
@@ -0,0 +1,104 @@
1comment "UBI debugging options"
2 depends on MTD_UBI
3
4config MTD_UBI_DEBUG
5 bool "UBI debugging"
6 depends on SYSFS
7 depends on MTD_UBI
8 select DEBUG_FS
9 select KALLSYMS_ALL
10 help
11 This option enables UBI debugging.
12
13config MTD_UBI_DEBUG_MSG
14 bool "UBI debugging messages"
15 depends on MTD_UBI_DEBUG
16 default n
17 help
18 This option enables UBI debugging messages.
19
20config MTD_UBI_DEBUG_PARANOID
21 bool "Extra self-checks"
22 default n
23 depends on MTD_UBI_DEBUG
24 help
25 This option enables extra checks in UBI code. Note this slows UBI down
26 significantly.
27
28config MTD_UBI_DEBUG_DISABLE_BGT
29 bool "Do not enable the UBI background thread"
30 depends on MTD_UBI_DEBUG
31 default n
32 help
33 This option switches the background thread off by default. The thread
34 may be also be enabled/disabled via UBI sysfs.
35
36config MTD_UBI_DEBUG_USERSPACE_IO
37 bool "Direct user-space write/erase support"
38 default n
39 depends on MTD_UBI_DEBUG
40 help
41 By default, users cannot directly write and erase individual
42 eraseblocks of dynamic volumes, and have to use update operation
43 instead. This option enables this capability - it is very useful for
44 debugging and testing.
45
46config MTD_UBI_DEBUG_EMULATE_BITFLIPS
47 bool "Emulate flash bit-flips"
48 depends on MTD_UBI_DEBUG
49 default n
50 help
51 This option emulates bit-flips with probability 1/50, which in turn
52 causes scrubbing. Useful for debugging and stressing UBI.
53
54config MTD_UBI_DEBUG_EMULATE_WRITE_FAILURES
55 bool "Emulate flash write failures"
56 depends on MTD_UBI_DEBUG
57 default n
58 help
59 This option emulates write failures with probability 1/100. Useful for
60 debugging and testing how UBI handlines errors.
61
62config MTD_UBI_DEBUG_EMULATE_ERASE_FAILURES
63 bool "Emulate flash erase failures"
64 depends on MTD_UBI_DEBUG
65 default n
66 help
67 This option emulates erase failures with probability 1/100. Useful for
68 debugging and testing how UBI handlines errors.
69
70menu "Additional UBI debugging messages"
71 depends on MTD_UBI_DEBUG
72
73config MTD_UBI_DEBUG_MSG_BLD
74 bool "Additional UBI initialization and build messages"
75 default n
76 depends on MTD_UBI_DEBUG
77 help
78 This option enables detailed UBI initialization and device build
79 debugging messages.
80
81config MTD_UBI_DEBUG_MSG_EBA
82 bool "Eraseblock association unit messages"
83 default n
84 depends on MTD_UBI_DEBUG
85 help
86 This option enables debugging messages from the UBI eraseblock
87 association unit.
88
89config MTD_UBI_DEBUG_MSG_WL
90 bool "Wear-leveling unit messages"
91 default n
92 depends on MTD_UBI_DEBUG
93 help
94 This option enables debugging messages from the UBI wear-leveling
95 unit.
96
97config MTD_UBI_DEBUG_MSG_IO
98 bool "Input/output unit messages"
99 default n
100 depends on MTD_UBI_DEBUG
101 help
102 This option enables debugging messages from the UBI input/output unit.
103
104endmenu # UBI debugging messages
diff --git a/drivers/mtd/ubi/Makefile b/drivers/mtd/ubi/Makefile
new file mode 100644
index 000000000000..dd834e04151b
--- /dev/null
+++ b/drivers/mtd/ubi/Makefile
@@ -0,0 +1,7 @@
1obj-$(CONFIG_MTD_UBI) += ubi.o
2
3ubi-y += vtbl.o vmt.o upd.o build.o cdev.o kapi.o eba.o io.o wl.o scan.o
4ubi-y += misc.o
5
6ubi-$(CONFIG_MTD_UBI_DEBUG) += debug.o
7ubi-$(CONFIG_MTD_UBI_GLUEBI) += gluebi.o
diff --git a/drivers/mtd/ubi/build.c b/drivers/mtd/ubi/build.c
new file mode 100644
index 000000000000..555d594d1811
--- /dev/null
+++ b/drivers/mtd/ubi/build.c
@@ -0,0 +1,848 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
3 * Copyright (c) Nokia Corporation, 2007
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 * the GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 *
19 * Author: Artem Bityutskiy (Битюцкий Артём),
20 * Frank Haverkamp
21 */
22
23/*
24 * This file includes UBI initialization and building of UBI devices. At the
25 * moment UBI devices may only be added while UBI is initialized, but dynamic
26 * device add/remove functionality is planned. Also, at the moment we only
27 * attach UBI devices by scanning, which will become a bottleneck when flashes
28 * reach certain large size. Then one may improve UBI and add other methods.
29 */
30
31#include <linux/err.h>
32#include <linux/module.h>
33#include <linux/moduleparam.h>
34#include <linux/stringify.h>
35#include <linux/stat.h>
36#include "ubi.h"
37
38/* Maximum length of the 'mtd=' parameter */
39#define MTD_PARAM_LEN_MAX 64
40
41/**
42 * struct mtd_dev_param - MTD device parameter description data structure.
43 * @name: MTD device name or number string
44 * @vid_hdr_offs: VID header offset
45 * @data_offs: data offset
46 */
47struct mtd_dev_param
48{
49 char name[MTD_PARAM_LEN_MAX];
50 int vid_hdr_offs;
51 int data_offs;
52};
53
54/* Numbers of elements set in the @mtd_dev_param array */
55static int mtd_devs = 0;
56
57/* MTD devices specification parameters */
58static struct mtd_dev_param mtd_dev_param[UBI_MAX_DEVICES];
59
60/* Number of UBI devices in system */
61int ubi_devices_cnt;
62
63/* All UBI devices in system */
64struct ubi_device *ubi_devices[UBI_MAX_DEVICES];
65
66/* Root UBI "class" object (corresponds to '/<sysfs>/class/ubi/') */
67struct class *ubi_class;
68
69/* "Show" method for files in '/<sysfs>/class/ubi/' */
70static ssize_t ubi_version_show(struct class *class, char *buf)
71{
72 return sprintf(buf, "%d\n", UBI_VERSION);
73}
74
75/* UBI version attribute ('/<sysfs>/class/ubi/version') */
76static struct class_attribute ubi_version =
77 __ATTR(version, S_IRUGO, ubi_version_show, NULL);
78
79static ssize_t dev_attribute_show(struct device *dev,
80 struct device_attribute *attr, char *buf);
81
82/* UBI device attributes (correspond to files in '/<sysfs>/class/ubi/ubiX') */
83static struct device_attribute dev_eraseblock_size =
84 __ATTR(eraseblock_size, S_IRUGO, dev_attribute_show, NULL);
85static struct device_attribute dev_avail_eraseblocks =
86 __ATTR(avail_eraseblocks, S_IRUGO, dev_attribute_show, NULL);
87static struct device_attribute dev_total_eraseblocks =
88 __ATTR(total_eraseblocks, S_IRUGO, dev_attribute_show, NULL);
89static struct device_attribute dev_volumes_count =
90 __ATTR(volumes_count, S_IRUGO, dev_attribute_show, NULL);
91static struct device_attribute dev_max_ec =
92 __ATTR(max_ec, S_IRUGO, dev_attribute_show, NULL);
93static struct device_attribute dev_reserved_for_bad =
94 __ATTR(reserved_for_bad, S_IRUGO, dev_attribute_show, NULL);
95static struct device_attribute dev_bad_peb_count =
96 __ATTR(bad_peb_count, S_IRUGO, dev_attribute_show, NULL);
97static struct device_attribute dev_max_vol_count =
98 __ATTR(max_vol_count, S_IRUGO, dev_attribute_show, NULL);
99static struct device_attribute dev_min_io_size =
100 __ATTR(min_io_size, S_IRUGO, dev_attribute_show, NULL);
101static struct device_attribute dev_bgt_enabled =
102 __ATTR(bgt_enabled, S_IRUGO, dev_attribute_show, NULL);
103
104/* "Show" method for files in '/<sysfs>/class/ubi/ubiX/' */
105static ssize_t dev_attribute_show(struct device *dev,
106 struct device_attribute *attr, char *buf)
107{
108 const struct ubi_device *ubi;
109
110 ubi = container_of(dev, struct ubi_device, dev);
111 if (attr == &dev_eraseblock_size)
112 return sprintf(buf, "%d\n", ubi->leb_size);
113 else if (attr == &dev_avail_eraseblocks)
114 return sprintf(buf, "%d\n", ubi->avail_pebs);
115 else if (attr == &dev_total_eraseblocks)
116 return sprintf(buf, "%d\n", ubi->good_peb_count);
117 else if (attr == &dev_volumes_count)
118 return sprintf(buf, "%d\n", ubi->vol_count);
119 else if (attr == &dev_max_ec)
120 return sprintf(buf, "%d\n", ubi->max_ec);
121 else if (attr == &dev_reserved_for_bad)
122 return sprintf(buf, "%d\n", ubi->beb_rsvd_pebs);
123 else if (attr == &dev_bad_peb_count)
124 return sprintf(buf, "%d\n", ubi->bad_peb_count);
125 else if (attr == &dev_max_vol_count)
126 return sprintf(buf, "%d\n", ubi->vtbl_slots);
127 else if (attr == &dev_min_io_size)
128 return sprintf(buf, "%d\n", ubi->min_io_size);
129 else if (attr == &dev_bgt_enabled)
130 return sprintf(buf, "%d\n", ubi->thread_enabled);
131 else
132 BUG();
133
134 return 0;
135}
136
137/* Fake "release" method for UBI devices */
138static void dev_release(struct device *dev) { }
139
140/**
141 * ubi_sysfs_init - initialize sysfs for an UBI device.
142 * @ubi: UBI device description object
143 *
144 * This function returns zero in case of success and a negative error code in
145 * case of failure.
146 */
147static int ubi_sysfs_init(struct ubi_device *ubi)
148{
149 int err;
150
151 ubi->dev.release = dev_release;
152 ubi->dev.devt = MKDEV(ubi->major, 0);
153 ubi->dev.class = ubi_class;
154 sprintf(&ubi->dev.bus_id[0], UBI_NAME_STR"%d", ubi->ubi_num);
155 err = device_register(&ubi->dev);
156 if (err)
157 goto out;
158
159 err = device_create_file(&ubi->dev, &dev_eraseblock_size);
160 if (err)
161 goto out_unregister;
162 err = device_create_file(&ubi->dev, &dev_avail_eraseblocks);
163 if (err)
164 goto out_eraseblock_size;
165 err = device_create_file(&ubi->dev, &dev_total_eraseblocks);
166 if (err)
167 goto out_avail_eraseblocks;
168 err = device_create_file(&ubi->dev, &dev_volumes_count);
169 if (err)
170 goto out_total_eraseblocks;
171 err = device_create_file(&ubi->dev, &dev_max_ec);
172 if (err)
173 goto out_volumes_count;
174 err = device_create_file(&ubi->dev, &dev_reserved_for_bad);
175 if (err)
176 goto out_volumes_max_ec;
177 err = device_create_file(&ubi->dev, &dev_bad_peb_count);
178 if (err)
179 goto out_reserved_for_bad;
180 err = device_create_file(&ubi->dev, &dev_max_vol_count);
181 if (err)
182 goto out_bad_peb_count;
183 err = device_create_file(&ubi->dev, &dev_min_io_size);
184 if (err)
185 goto out_max_vol_count;
186 err = device_create_file(&ubi->dev, &dev_bgt_enabled);
187 if (err)
188 goto out_min_io_size;
189
190 return 0;
191
192out_min_io_size:
193 device_remove_file(&ubi->dev, &dev_min_io_size);
194out_max_vol_count:
195 device_remove_file(&ubi->dev, &dev_max_vol_count);
196out_bad_peb_count:
197 device_remove_file(&ubi->dev, &dev_bad_peb_count);
198out_reserved_for_bad:
199 device_remove_file(&ubi->dev, &dev_reserved_for_bad);
200out_volumes_max_ec:
201 device_remove_file(&ubi->dev, &dev_max_ec);
202out_volumes_count:
203 device_remove_file(&ubi->dev, &dev_volumes_count);
204out_total_eraseblocks:
205 device_remove_file(&ubi->dev, &dev_total_eraseblocks);
206out_avail_eraseblocks:
207 device_remove_file(&ubi->dev, &dev_avail_eraseblocks);
208out_eraseblock_size:
209 device_remove_file(&ubi->dev, &dev_eraseblock_size);
210out_unregister:
211 device_unregister(&ubi->dev);
212out:
213 ubi_err("failed to initialize sysfs for %s", ubi->ubi_name);
214 return err;
215}
216
217/**
218 * ubi_sysfs_close - close sysfs for an UBI device.
219 * @ubi: UBI device description object
220 */
221static void ubi_sysfs_close(struct ubi_device *ubi)
222{
223 device_remove_file(&ubi->dev, &dev_bgt_enabled);
224 device_remove_file(&ubi->dev, &dev_min_io_size);
225 device_remove_file(&ubi->dev, &dev_max_vol_count);
226 device_remove_file(&ubi->dev, &dev_bad_peb_count);
227 device_remove_file(&ubi->dev, &dev_reserved_for_bad);
228 device_remove_file(&ubi->dev, &dev_max_ec);
229 device_remove_file(&ubi->dev, &dev_volumes_count);
230 device_remove_file(&ubi->dev, &dev_total_eraseblocks);
231 device_remove_file(&ubi->dev, &dev_avail_eraseblocks);
232 device_remove_file(&ubi->dev, &dev_eraseblock_size);
233 device_unregister(&ubi->dev);
234}
235
236/**
237 * kill_volumes - destroy all volumes.
238 * @ubi: UBI device description object
239 */
240static void kill_volumes(struct ubi_device *ubi)
241{
242 int i;
243
244 for (i = 0; i < ubi->vtbl_slots; i++)
245 if (ubi->volumes[i])
246 ubi_free_volume(ubi, i);
247}
248
249/**
250 * uif_init - initialize user interfaces for an UBI device.
251 * @ubi: UBI device description object
252 *
253 * This function returns zero in case of success and a negative error code in
254 * case of failure.
255 */
256static int uif_init(struct ubi_device *ubi)
257{
258 int i, err;
259 dev_t dev;
260
261 mutex_init(&ubi->vtbl_mutex);
262 spin_lock_init(&ubi->volumes_lock);
263
264 sprintf(ubi->ubi_name, UBI_NAME_STR "%d", ubi->ubi_num);
265
266 /*
267 * Major numbers for the UBI character devices are allocated
268 * dynamically. Major numbers of volume character devices are
269 * equivalent to ones of the corresponding UBI character device. Minor
270 * numbers of UBI character devices are 0, while minor numbers of
271 * volume character devices start from 1. Thus, we allocate one major
272 * number and ubi->vtbl_slots + 1 minor numbers.
273 */
274 err = alloc_chrdev_region(&dev, 0, ubi->vtbl_slots + 1, ubi->ubi_name);
275 if (err) {
276 ubi_err("cannot register UBI character devices");
277 return err;
278 }
279
280 cdev_init(&ubi->cdev, &ubi_cdev_operations);
281 ubi->major = MAJOR(dev);
282 dbg_msg("%s major is %u", ubi->ubi_name, ubi->major);
283 ubi->cdev.owner = THIS_MODULE;
284
285 dev = MKDEV(ubi->major, 0);
286 err = cdev_add(&ubi->cdev, dev, 1);
287 if (err) {
288 ubi_err("cannot add character device %s", ubi->ubi_name);
289 goto out_unreg;
290 }
291
292 err = ubi_sysfs_init(ubi);
293 if (err)
294 goto out_cdev;
295
296 for (i = 0; i < ubi->vtbl_slots; i++)
297 if (ubi->volumes[i]) {
298 err = ubi_add_volume(ubi, i);
299 if (err)
300 goto out_volumes;
301 }
302
303 return 0;
304
305out_volumes:
306 kill_volumes(ubi);
307 ubi_sysfs_close(ubi);
308out_cdev:
309 cdev_del(&ubi->cdev);
310out_unreg:
311 unregister_chrdev_region(MKDEV(ubi->major, 0),
312 ubi->vtbl_slots + 1);
313 return err;
314}
315
316/**
317 * uif_close - close user interfaces for an UBI device.
318 * @ubi: UBI device description object
319 */
320static void uif_close(struct ubi_device *ubi)
321{
322 kill_volumes(ubi);
323 ubi_sysfs_close(ubi);
324 cdev_del(&ubi->cdev);
325 unregister_chrdev_region(MKDEV(ubi->major, 0), ubi->vtbl_slots + 1);
326}
327
328/**
329 * attach_by_scanning - attach an MTD device using scanning method.
330 * @ubi: UBI device descriptor
331 *
332 * This function returns zero in case of success and a negative error code in
333 * case of failure.
334 *
335 * Note, currently this is the only method to attach UBI devices. Hopefully in
336 * the future we'll have more scalable attaching methods and avoid full media
337 * scanning. But even in this case scanning will be needed as a fall-back
338 * attaching method if there are some on-flash table corruptions.
339 */
340static int attach_by_scanning(struct ubi_device *ubi)
341{
342 int err;
343 struct ubi_scan_info *si;
344
345 si = ubi_scan(ubi);
346 if (IS_ERR(si))
347 return PTR_ERR(si);
348
349 ubi->bad_peb_count = si->bad_peb_count;
350 ubi->good_peb_count = ubi->peb_count - ubi->bad_peb_count;
351 ubi->max_ec = si->max_ec;
352 ubi->mean_ec = si->mean_ec;
353
354 err = ubi_read_volume_table(ubi, si);
355 if (err)
356 goto out_si;
357
358 err = ubi_wl_init_scan(ubi, si);
359 if (err)
360 goto out_vtbl;
361
362 err = ubi_eba_init_scan(ubi, si);
363 if (err)
364 goto out_wl;
365
366 ubi_scan_destroy_si(si);
367 return 0;
368
369out_wl:
370 ubi_wl_close(ubi);
371out_vtbl:
372 kfree(ubi->vtbl);
373out_si:
374 ubi_scan_destroy_si(si);
375 return err;
376}
377
378/**
379 * io_init - initialize I/O unit for a given UBI device.
380 * @ubi: UBI device description object
381 *
382 * If @ubi->vid_hdr_offset or @ubi->leb_start is zero, default offsets are
383 * assumed:
384 * o EC header is always at offset zero - this cannot be changed;
385 * o VID header starts just after the EC header at the closest address
386 * aligned to @io->@hdrs_min_io_size;
387 * o data starts just after the VID header at the closest address aligned to
388 * @io->@min_io_size
389 *
390 * This function returns zero in case of success and a negative error code in
391 * case of failure.
392 */
393static int io_init(struct ubi_device *ubi)
394{
395 if (ubi->mtd->numeraseregions != 0) {
396 /*
397 * Some flashes have several erase regions. Different regions
398 * may have different eraseblock size and other
399 * characteristics. It looks like mostly multi-region flashes
400 * have one "main" region and one or more small regions to
401 * store boot loader code or boot parameters or whatever. I
402 * guess we should just pick the largest region. But this is
403 * not implemented.
404 */
405 ubi_err("multiple regions, not implemented");
406 return -EINVAL;
407 }
408
409 /*
410 * Note, in this implementation we support MTD devices with 0x7FFFFFFF
411 * physical eraseblocks maximum.
412 */
413
414 ubi->peb_size = ubi->mtd->erasesize;
415 ubi->peb_count = ubi->mtd->size / ubi->mtd->erasesize;
416 ubi->flash_size = ubi->mtd->size;
417
418 if (ubi->mtd->block_isbad && ubi->mtd->block_markbad)
419 ubi->bad_allowed = 1;
420
421 ubi->min_io_size = ubi->mtd->writesize;
422 ubi->hdrs_min_io_size = ubi->mtd->writesize >> ubi->mtd->subpage_sft;
423
424 /* Make sure minimal I/O unit is power of 2 */
425 if (ubi->min_io_size == 0 ||
426 (ubi->min_io_size & (ubi->min_io_size - 1))) {
427 ubi_err("bad min. I/O unit");
428 return -EINVAL;
429 }
430
431 ubi_assert(ubi->hdrs_min_io_size > 0);
432 ubi_assert(ubi->hdrs_min_io_size <= ubi->min_io_size);
433 ubi_assert(ubi->min_io_size % ubi->hdrs_min_io_size == 0);
434
435 /* Calculate default aligned sizes of EC and VID headers */
436 ubi->ec_hdr_alsize = ALIGN(UBI_EC_HDR_SIZE, ubi->hdrs_min_io_size);
437 ubi->vid_hdr_alsize = ALIGN(UBI_VID_HDR_SIZE, ubi->hdrs_min_io_size);
438
439 dbg_msg("min_io_size %d", ubi->min_io_size);
440 dbg_msg("hdrs_min_io_size %d", ubi->hdrs_min_io_size);
441 dbg_msg("ec_hdr_alsize %d", ubi->ec_hdr_alsize);
442 dbg_msg("vid_hdr_alsize %d", ubi->vid_hdr_alsize);
443
444 if (ubi->vid_hdr_offset == 0)
445 /* Default offset */
446 ubi->vid_hdr_offset = ubi->vid_hdr_aloffset =
447 ubi->ec_hdr_alsize;
448 else {
449 ubi->vid_hdr_aloffset = ubi->vid_hdr_offset &
450 ~(ubi->hdrs_min_io_size - 1);
451 ubi->vid_hdr_shift = ubi->vid_hdr_offset -
452 ubi->vid_hdr_aloffset;
453 }
454
455 /* Similar for the data offset */
456 if (ubi->leb_start == 0) {
457 ubi->leb_start = ubi->vid_hdr_offset + ubi->vid_hdr_alsize;
458 ubi->leb_start = ALIGN(ubi->leb_start, ubi->min_io_size);
459 }
460
461 dbg_msg("vid_hdr_offset %d", ubi->vid_hdr_offset);
462 dbg_msg("vid_hdr_aloffset %d", ubi->vid_hdr_aloffset);
463 dbg_msg("vid_hdr_shift %d", ubi->vid_hdr_shift);
464 dbg_msg("leb_start %d", ubi->leb_start);
465
466 /* The shift must be aligned to 32-bit boundary */
467 if (ubi->vid_hdr_shift % 4) {
468 ubi_err("unaligned VID header shift %d",
469 ubi->vid_hdr_shift);
470 return -EINVAL;
471 }
472
473 /* Check sanity */
474 if (ubi->vid_hdr_offset < UBI_EC_HDR_SIZE ||
475 ubi->leb_start < ubi->vid_hdr_offset + UBI_VID_HDR_SIZE ||
476 ubi->leb_start > ubi->peb_size - UBI_VID_HDR_SIZE ||
477 ubi->leb_start % ubi->min_io_size) {
478 ubi_err("bad VID header (%d) or data offsets (%d)",
479 ubi->vid_hdr_offset, ubi->leb_start);
480 return -EINVAL;
481 }
482
483 /*
484 * It may happen that EC and VID headers are situated in one minimal
485 * I/O unit. In this case we can only accept this UBI image in
486 * read-only mode.
487 */
488 if (ubi->vid_hdr_offset + UBI_VID_HDR_SIZE <= ubi->hdrs_min_io_size) {
489 ubi_warn("EC and VID headers are in the same minimal I/O unit, "
490 "switch to read-only mode");
491 ubi->ro_mode = 1;
492 }
493
494 ubi->leb_size = ubi->peb_size - ubi->leb_start;
495
496 if (!(ubi->mtd->flags & MTD_WRITEABLE)) {
497 ubi_msg("MTD device %d is write-protected, attach in "
498 "read-only mode", ubi->mtd->index);
499 ubi->ro_mode = 1;
500 }
501
502 dbg_msg("leb_size %d", ubi->leb_size);
503 dbg_msg("ro_mode %d", ubi->ro_mode);
504
505 /*
506 * Note, ideally, we have to initialize ubi->bad_peb_count here. But
507 * unfortunately, MTD does not provide this information. We should loop
508 * over all physical eraseblocks and invoke mtd->block_is_bad() for
509 * each physical eraseblock. So, we skip ubi->bad_peb_count
510 * uninitialized and initialize it after scanning.
511 */
512
513 return 0;
514}
515
516/**
517 * attach_mtd_dev - attach an MTD device.
518 * @mtd_dev: MTD device name or number string
519 * @vid_hdr_offset: VID header offset
520 * @data_offset: data offset
521 *
522 * This function attaches an MTD device to UBI. It first treats @mtd_dev as the
523 * MTD device name, and tries to open it by this name. If it is unable to open,
524 * it tries to convert @mtd_dev to an integer and open the MTD device by its
525 * number. Returns zero in case of success and a negative error code in case of
526 * failure.
527 */
528static int attach_mtd_dev(const char *mtd_dev, int vid_hdr_offset,
529 int data_offset)
530{
531 struct ubi_device *ubi;
532 struct mtd_info *mtd;
533 int i, err;
534
535 mtd = get_mtd_device_nm(mtd_dev);
536 if (IS_ERR(mtd)) {
537 int mtd_num;
538 char *endp;
539
540 if (PTR_ERR(mtd) != -ENODEV)
541 return PTR_ERR(mtd);
542
543 /*
544 * Probably this is not MTD device name but MTD device number -
545 * check this out.
546 */
547 mtd_num = simple_strtoul(mtd_dev, &endp, 0);
548 if (*endp != '\0' || mtd_dev == endp) {
549 ubi_err("incorrect MTD device: \"%s\"", mtd_dev);
550 return -ENODEV;
551 }
552
553 mtd = get_mtd_device(NULL, mtd_num);
554 if (IS_ERR(mtd))
555 return PTR_ERR(mtd);
556 }
557
558 /* Check if we already have the same MTD device attached */
559 for (i = 0; i < ubi_devices_cnt; i++)
560 if (ubi_devices[i]->mtd->index == mtd->index) {
561 ubi_err("mtd%d is already attached to ubi%d",
562 mtd->index, i);
563 err = -EINVAL;
564 goto out_mtd;
565 }
566
567 ubi = ubi_devices[ubi_devices_cnt] = kzalloc(sizeof(struct ubi_device),
568 GFP_KERNEL);
569 if (!ubi) {
570 err = -ENOMEM;
571 goto out_mtd;
572 }
573
574 ubi->ubi_num = ubi_devices_cnt;
575 ubi->mtd = mtd;
576
577 dbg_msg("attaching mtd%d to ubi%d: VID header offset %d data offset %d",
578 ubi->mtd->index, ubi_devices_cnt, vid_hdr_offset, data_offset);
579
580 ubi->vid_hdr_offset = vid_hdr_offset;
581 ubi->leb_start = data_offset;
582 err = io_init(ubi);
583 if (err)
584 goto out_free;
585
586 err = attach_by_scanning(ubi);
587 if (err) {
588 dbg_err("failed to attach by scanning, error %d", err);
589 goto out_free;
590 }
591
592 err = uif_init(ubi);
593 if (err)
594 goto out_detach;
595
596 ubi_devices_cnt += 1;
597
598 ubi_msg("attached mtd%d to ubi%d", ubi->mtd->index, ubi_devices_cnt);
599 ubi_msg("MTD device name: \"%s\"", ubi->mtd->name);
600 ubi_msg("MTD device size: %llu MiB", ubi->flash_size >> 20);
601 ubi_msg("physical eraseblock size: %d bytes (%d KiB)",
602 ubi->peb_size, ubi->peb_size >> 10);
603 ubi_msg("logical eraseblock size: %d bytes", ubi->leb_size);
604 ubi_msg("number of good PEBs: %d", ubi->good_peb_count);
605 ubi_msg("number of bad PEBs: %d", ubi->bad_peb_count);
606 ubi_msg("smallest flash I/O unit: %d", ubi->min_io_size);
607 ubi_msg("VID header offset: %d (aligned %d)",
608 ubi->vid_hdr_offset, ubi->vid_hdr_aloffset);
609 ubi_msg("data offset: %d", ubi->leb_start);
610 ubi_msg("max. allowed volumes: %d", ubi->vtbl_slots);
611 ubi_msg("wear-leveling threshold: %d", CONFIG_MTD_UBI_WL_THRESHOLD);
612 ubi_msg("number of internal volumes: %d", UBI_INT_VOL_COUNT);
613 ubi_msg("number of user volumes: %d",
614 ubi->vol_count - UBI_INT_VOL_COUNT);
615 ubi_msg("available PEBs: %d", ubi->avail_pebs);
616 ubi_msg("total number of reserved PEBs: %d", ubi->rsvd_pebs);
617 ubi_msg("number of PEBs reserved for bad PEB handling: %d",
618 ubi->beb_rsvd_pebs);
619 ubi_msg("max/mean erase counter: %d/%d", ubi->max_ec, ubi->mean_ec);
620
621 /* Enable the background thread */
622 if (!DBG_DISABLE_BGT) {
623 ubi->thread_enabled = 1;
624 wake_up_process(ubi->bgt_thread);
625 }
626
627 return 0;
628
629out_detach:
630 ubi_eba_close(ubi);
631 ubi_wl_close(ubi);
632 kfree(ubi->vtbl);
633out_free:
634 kfree(ubi);
635out_mtd:
636 put_mtd_device(mtd);
637 ubi_devices[ubi_devices_cnt] = NULL;
638 return err;
639}
640
641/**
642 * detach_mtd_dev - detach an MTD device.
643 * @ubi: UBI device description object
644 */
645static void detach_mtd_dev(struct ubi_device *ubi)
646{
647 int ubi_num = ubi->ubi_num, mtd_num = ubi->mtd->index;
648
649 dbg_msg("detaching mtd%d from ubi%d", ubi->mtd->index, ubi_num);
650 uif_close(ubi);
651 ubi_eba_close(ubi);
652 ubi_wl_close(ubi);
653 kfree(ubi->vtbl);
654 put_mtd_device(ubi->mtd);
655 kfree(ubi_devices[ubi_num]);
656 ubi_devices[ubi_num] = NULL;
657 ubi_devices_cnt -= 1;
658 ubi_assert(ubi_devices_cnt >= 0);
659 ubi_msg("mtd%d is detached from ubi%d", mtd_num, ubi_num);
660}
661
662static int __init ubi_init(void)
663{
664 int err, i, k;
665
666 /* Ensure that EC and VID headers have correct size */
667 BUILD_BUG_ON(sizeof(struct ubi_ec_hdr) != 64);
668 BUILD_BUG_ON(sizeof(struct ubi_vid_hdr) != 64);
669
670 if (mtd_devs > UBI_MAX_DEVICES) {
671 printk("UBI error: too many MTD devices, maximum is %d\n",
672 UBI_MAX_DEVICES);
673 return -EINVAL;
674 }
675
676 ubi_class = class_create(THIS_MODULE, UBI_NAME_STR);
677 if (IS_ERR(ubi_class))
678 return PTR_ERR(ubi_class);
679
680 err = class_create_file(ubi_class, &ubi_version);
681 if (err)
682 goto out_class;
683
684 /* Attach MTD devices */
685 for (i = 0; i < mtd_devs; i++) {
686 struct mtd_dev_param *p = &mtd_dev_param[i];
687
688 cond_resched();
689
690 if (!p->name) {
691 dbg_err("empty name");
692 err = -EINVAL;
693 goto out_detach;
694 }
695
696 err = attach_mtd_dev(p->name, p->vid_hdr_offs, p->data_offs);
697 if (err)
698 goto out_detach;
699 }
700
701 return 0;
702
703out_detach:
704 for (k = 0; k < i; k++)
705 detach_mtd_dev(ubi_devices[k]);
706 class_remove_file(ubi_class, &ubi_version);
707out_class:
708 class_destroy(ubi_class);
709 return err;
710}
711module_init(ubi_init);
712
713static void __exit ubi_exit(void)
714{
715 int i, n = ubi_devices_cnt;
716
717 for (i = 0; i < n; i++)
718 detach_mtd_dev(ubi_devices[i]);
719 class_remove_file(ubi_class, &ubi_version);
720 class_destroy(ubi_class);
721}
722module_exit(ubi_exit);
723
724/**
725 * bytes_str_to_int - convert a string representing number of bytes to an
726 * integer.
727 * @str: the string to convert
728 *
729 * This function returns positive resulting integer in case of success and a
730 * negative error code in case of failure.
731 */
732static int __init bytes_str_to_int(const char *str)
733{
734 char *endp;
735 unsigned long result;
736
737 result = simple_strtoul(str, &endp, 0);
738 if (str == endp || result < 0) {
739 printk("UBI error: incorrect bytes count: \"%s\"\n", str);
740 return -EINVAL;
741 }
742
743 switch (*endp) {
744 case 'G':
745 result *= 1024;
746 case 'M':
747 result *= 1024;
748 case 'K':
749 case 'k':
750 result *= 1024;
751 if (endp[1] == 'i' && (endp[2] == '\0' ||
752 endp[2] == 'B' || endp[2] == 'b'))
753 endp += 2;
754 case '\0':
755 break;
756 default:
757 printk("UBI error: incorrect bytes count: \"%s\"\n", str);
758 return -EINVAL;
759 }
760
761 return result;
762}
763
764/**
765 * ubi_mtd_param_parse - parse the 'mtd=' UBI parameter.
766 * @val: the parameter value to parse
767 * @kp: not used
768 *
769 * This function returns zero in case of success and a negative error code in
770 * case of error.
771 */
772static int __init ubi_mtd_param_parse(const char *val, struct kernel_param *kp)
773{
774 int i, len;
775 struct mtd_dev_param *p;
776 char buf[MTD_PARAM_LEN_MAX];
777 char *pbuf = &buf[0];
778 char *tokens[3] = {NULL, NULL, NULL};
779
780 if (mtd_devs == UBI_MAX_DEVICES) {
781 printk("UBI error: too many parameters, max. is %d\n",
782 UBI_MAX_DEVICES);
783 return -EINVAL;
784 }
785
786 len = strnlen(val, MTD_PARAM_LEN_MAX);
787 if (len == MTD_PARAM_LEN_MAX) {
788 printk("UBI error: parameter \"%s\" is too long, max. is %d\n",
789 val, MTD_PARAM_LEN_MAX);
790 return -EINVAL;
791 }
792
793 if (len == 0) {
794 printk("UBI warning: empty 'mtd=' parameter - ignored\n");
795 return 0;
796 }
797
798 strcpy(buf, val);
799
800 /* Get rid of the final newline */
801 if (buf[len - 1] == '\n')
802 buf[len - 1] = 0;
803
804 for (i = 0; i < 3; i++)
805 tokens[i] = strsep(&pbuf, ",");
806
807 if (pbuf) {
808 printk("UBI error: too many arguments at \"%s\"\n", val);
809 return -EINVAL;
810 }
811
812 if (tokens[0] == '\0')
813 return -EINVAL;
814
815 p = &mtd_dev_param[mtd_devs];
816 strcpy(&p->name[0], tokens[0]);
817
818 if (tokens[1])
819 p->vid_hdr_offs = bytes_str_to_int(tokens[1]);
820 if (tokens[2])
821 p->data_offs = bytes_str_to_int(tokens[2]);
822
823 if (p->vid_hdr_offs < 0)
824 return p->vid_hdr_offs;
825 if (p->data_offs < 0)
826 return p->data_offs;
827
828 mtd_devs += 1;
829 return 0;
830}
831
832module_param_call(mtd, ubi_mtd_param_parse, NULL, NULL, 000);
833MODULE_PARM_DESC(mtd, "MTD devices to attach. Parameter format: "
834 "mtd=<name|num>[,<vid_hdr_offs>,<data_offs>]. "
835 "Multiple \"mtd\" parameters may be specified.\n"
836 "MTD devices may be specified by their number or name. "
837 "Optional \"vid_hdr_offs\" and \"data_offs\" parameters "
838 "specify UBI VID header position and data starting "
839 "position to be used by UBI.\n"
840 "Example: mtd=content,1984,2048 mtd=4 - attach MTD device"
841 "with name content using VID header offset 1984 and data "
842 "start 2048, and MTD device number 4 using default "
843 "offsets");
844
845MODULE_VERSION(__stringify(UBI_VERSION));
846MODULE_DESCRIPTION("UBI - Unsorted Block Images");
847MODULE_AUTHOR("Artem Bityutskiy");
848MODULE_LICENSE("GPL");
diff --git a/drivers/mtd/ubi/cdev.c b/drivers/mtd/ubi/cdev.c
new file mode 100644
index 000000000000..6612eb79bf17
--- /dev/null
+++ b/drivers/mtd/ubi/cdev.c
@@ -0,0 +1,722 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/*
22 * This file includes implementation of UBI character device operations.
23 *
24 * There are two kinds of character devices in UBI: UBI character devices and
25 * UBI volume character devices. UBI character devices allow users to
26 * manipulate whole volumes: create, remove, and re-size them. Volume character
27 * devices provide volume I/O capabilities.
28 *
29 * Major and minor numbers are assigned dynamically to both UBI and volume
30 * character devices.
31 */
32
33#include <linux/module.h>
34#include <linux/stat.h>
35#include <linux/ioctl.h>
36#include <linux/capability.h>
37#include <mtd/ubi-user.h>
38#include <asm/uaccess.h>
39#include <asm/div64.h>
40#include "ubi.h"
41
42/*
43 * Maximum sequence numbers of UBI and volume character device IOCTLs (direct
44 * logical eraseblock erase is a debug-only feature).
45 */
46#define UBI_CDEV_IOC_MAX_SEQ 2
47#ifndef CONFIG_MTD_UBI_DEBUG_USERSPACE_IO
48#define VOL_CDEV_IOC_MAX_SEQ 1
49#else
50#define VOL_CDEV_IOC_MAX_SEQ 2
51#endif
52
53/**
54 * major_to_device - get UBI device object by character device major number.
55 * @major: major number
56 *
57 * This function returns a pointer to the UBI device object.
58 */
59static struct ubi_device *major_to_device(int major)
60{
61 int i;
62
63 for (i = 0; i < ubi_devices_cnt; i++)
64 if (ubi_devices[i] && ubi_devices[i]->major == major)
65 return ubi_devices[i];
66 BUG();
67}
68
69/**
70 * get_exclusive - get exclusive access to an UBI volume.
71 * @desc: volume descriptor
72 *
73 * This function changes UBI volume open mode to "exclusive". Returns previous
74 * mode value (positive integer) in case of success and a negative error code
75 * in case of failure.
76 */
77static int get_exclusive(struct ubi_volume_desc *desc)
78{
79 int users, err;
80 struct ubi_volume *vol = desc->vol;
81
82 spin_lock(&vol->ubi->volumes_lock);
83 users = vol->readers + vol->writers + vol->exclusive;
84 ubi_assert(users > 0);
85 if (users > 1) {
86 dbg_err("%d users for volume %d", users, vol->vol_id);
87 err = -EBUSY;
88 } else {
89 vol->readers = vol->writers = 0;
90 vol->exclusive = 1;
91 err = desc->mode;
92 desc->mode = UBI_EXCLUSIVE;
93 }
94 spin_unlock(&vol->ubi->volumes_lock);
95
96 return err;
97}
98
99/**
100 * revoke_exclusive - revoke exclusive mode.
101 * @desc: volume descriptor
102 * @mode: new mode to switch to
103 */
104static void revoke_exclusive(struct ubi_volume_desc *desc, int mode)
105{
106 struct ubi_volume *vol = desc->vol;
107
108 spin_lock(&vol->ubi->volumes_lock);
109 ubi_assert(vol->readers == 0 && vol->writers == 0);
110 ubi_assert(vol->exclusive == 1 && desc->mode == UBI_EXCLUSIVE);
111 vol->exclusive = 0;
112 if (mode == UBI_READONLY)
113 vol->readers = 1;
114 else if (mode == UBI_READWRITE)
115 vol->writers = 1;
116 else
117 vol->exclusive = 1;
118 spin_unlock(&vol->ubi->volumes_lock);
119
120 desc->mode = mode;
121}
122
123static int vol_cdev_open(struct inode *inode, struct file *file)
124{
125 struct ubi_volume_desc *desc;
126 const struct ubi_device *ubi = major_to_device(imajor(inode));
127 int vol_id = iminor(inode) - 1;
128 int mode;
129
130 if (file->f_mode & FMODE_WRITE)
131 mode = UBI_READWRITE;
132 else
133 mode = UBI_READONLY;
134
135 dbg_msg("open volume %d, mode %d", vol_id, mode);
136
137 desc = ubi_open_volume(ubi->ubi_num, vol_id, mode);
138 if (IS_ERR(desc))
139 return PTR_ERR(desc);
140
141 file->private_data = desc;
142 return 0;
143}
144
145static int vol_cdev_release(struct inode *inode, struct file *file)
146{
147 struct ubi_volume_desc *desc = file->private_data;
148 struct ubi_volume *vol = desc->vol;
149
150 dbg_msg("release volume %d, mode %d", vol->vol_id, desc->mode);
151
152 if (vol->updating) {
153 ubi_warn("update of volume %d not finished, volume is damaged",
154 vol->vol_id);
155 vol->updating = 0;
156 kfree(vol->upd_buf);
157 }
158
159 ubi_close_volume(desc);
160 return 0;
161}
162
163static loff_t vol_cdev_llseek(struct file *file, loff_t offset, int origin)
164{
165 struct ubi_volume_desc *desc = file->private_data;
166 struct ubi_volume *vol = desc->vol;
167 loff_t new_offset;
168
169 if (vol->updating) {
170 /* Update is in progress, seeking is prohibited */
171 dbg_err("updating");
172 return -EBUSY;
173 }
174
175 switch (origin) {
176 case 0: /* SEEK_SET */
177 new_offset = offset;
178 break;
179 case 1: /* SEEK_CUR */
180 new_offset = file->f_pos + offset;
181 break;
182 case 2: /* SEEK_END */
183 new_offset = vol->used_bytes + offset;
184 break;
185 default:
186 return -EINVAL;
187 }
188
189 if (new_offset < 0 || new_offset > vol->used_bytes) {
190 dbg_err("bad seek %lld", new_offset);
191 return -EINVAL;
192 }
193
194 dbg_msg("seek volume %d, offset %lld, origin %d, new offset %lld",
195 vol->vol_id, offset, origin, new_offset);
196
197 file->f_pos = new_offset;
198 return new_offset;
199}
200
201static ssize_t vol_cdev_read(struct file *file, __user char *buf, size_t count,
202 loff_t *offp)
203{
204 struct ubi_volume_desc *desc = file->private_data;
205 struct ubi_volume *vol = desc->vol;
206 struct ubi_device *ubi = vol->ubi;
207 int err, lnum, off, len, vol_id = desc->vol->vol_id, tbuf_size;
208 size_t count_save = count;
209 void *tbuf;
210 uint64_t tmp;
211
212 dbg_msg("read %zd bytes from offset %lld of volume %d",
213 count, *offp, vol_id);
214
215 if (vol->updating) {
216 dbg_err("updating");
217 return -EBUSY;
218 }
219 if (vol->upd_marker) {
220 dbg_err("damaged volume, update marker is set");
221 return -EBADF;
222 }
223 if (*offp == vol->used_bytes || count == 0)
224 return 0;
225
226 if (vol->corrupted)
227 dbg_msg("read from corrupted volume %d", vol_id);
228
229 if (*offp + count > vol->used_bytes)
230 count_save = count = vol->used_bytes - *offp;
231
232 tbuf_size = vol->usable_leb_size;
233 if (count < tbuf_size)
234 tbuf_size = ALIGN(count, ubi->min_io_size);
235 tbuf = kmalloc(tbuf_size, GFP_KERNEL);
236 if (!tbuf)
237 return -ENOMEM;
238
239 len = count > tbuf_size ? tbuf_size : count;
240
241 tmp = *offp;
242 off = do_div(tmp, vol->usable_leb_size);
243 lnum = tmp;
244
245 do {
246 cond_resched();
247
248 if (off + len >= vol->usable_leb_size)
249 len = vol->usable_leb_size - off;
250
251 err = ubi_eba_read_leb(ubi, vol_id, lnum, tbuf, off, len, 0);
252 if (err)
253 break;
254
255 off += len;
256 if (off == vol->usable_leb_size) {
257 lnum += 1;
258 off -= vol->usable_leb_size;
259 }
260
261 count -= len;
262 *offp += len;
263
264 err = copy_to_user(buf, tbuf, len);
265 if (err) {
266 err = -EFAULT;
267 break;
268 }
269
270 buf += len;
271 len = count > tbuf_size ? tbuf_size : count;
272 } while (count);
273
274 kfree(tbuf);
275 return err ? err : count_save - count;
276}
277
278#ifdef CONFIG_MTD_UBI_DEBUG_USERSPACE_IO
279
280/*
281 * This function allows to directly write to dynamic UBI volumes, without
282 * issuing the volume update operation. Available only as a debugging feature.
283 * Very useful for testing UBI.
284 */
285static ssize_t vol_cdev_direct_write(struct file *file, const char __user *buf,
286 size_t count, loff_t *offp)
287{
288 struct ubi_volume_desc *desc = file->private_data;
289 struct ubi_volume *vol = desc->vol;
290 struct ubi_device *ubi = vol->ubi;
291 int lnum, off, len, tbuf_size, vol_id = vol->vol_id, err = 0;
292 size_t count_save = count;
293 char *tbuf;
294 uint64_t tmp;
295
296 dbg_msg("requested: write %zd bytes to offset %lld of volume %u",
297 count, *offp, desc->vol->vol_id);
298
299 if (vol->vol_type == UBI_STATIC_VOLUME)
300 return -EROFS;
301
302 tmp = *offp;
303 off = do_div(tmp, vol->usable_leb_size);
304 lnum = tmp;
305
306 if (off % ubi->min_io_size) {
307 dbg_err("unaligned position");
308 return -EINVAL;
309 }
310
311 if (*offp + count > vol->used_bytes)
312 count_save = count = vol->used_bytes - *offp;
313
314 /* We can write only in fractions of the minimum I/O unit */
315 if (count % ubi->min_io_size) {
316 dbg_err("unaligned write length");
317 return -EINVAL;
318 }
319
320 tbuf_size = vol->usable_leb_size;
321 if (count < tbuf_size)
322 tbuf_size = ALIGN(count, ubi->min_io_size);
323 tbuf = kmalloc(tbuf_size, GFP_KERNEL);
324 if (!tbuf)
325 return -ENOMEM;
326
327 len = count > tbuf_size ? tbuf_size : count;
328
329 while (count) {
330 cond_resched();
331
332 if (off + len >= vol->usable_leb_size)
333 len = vol->usable_leb_size - off;
334
335 err = copy_from_user(tbuf, buf, len);
336 if (err) {
337 err = -EFAULT;
338 break;
339 }
340
341 err = ubi_eba_write_leb(ubi, vol_id, lnum, tbuf, off, len,
342 UBI_UNKNOWN);
343 if (err)
344 break;
345
346 off += len;
347 if (off == vol->usable_leb_size) {
348 lnum += 1;
349 off -= vol->usable_leb_size;
350 }
351
352 count -= len;
353 *offp += len;
354 buf += len;
355 len = count > tbuf_size ? tbuf_size : count;
356 }
357
358 kfree(tbuf);
359 return err ? err : count_save - count;
360}
361
362#else
363#define vol_cdev_direct_write(file, buf, count, offp) -EPERM
364#endif /* CONFIG_MTD_UBI_DEBUG_USERSPACE_IO */
365
366static ssize_t vol_cdev_write(struct file *file, const char __user *buf,
367 size_t count, loff_t *offp)
368{
369 int err = 0;
370 struct ubi_volume_desc *desc = file->private_data;
371 struct ubi_volume *vol = desc->vol;
372 struct ubi_device *ubi = vol->ubi;
373
374 if (!vol->updating)
375 return vol_cdev_direct_write(file, buf, count, offp);
376
377 err = ubi_more_update_data(ubi, vol->vol_id, buf, count);
378 if (err < 0) {
379 ubi_err("cannot write %zd bytes of update data", count);
380 return err;
381 }
382
383 if (err) {
384 /*
385 * Update is finished, @err contains number of actually written
386 * bytes now.
387 */
388 count = err;
389
390 err = ubi_check_volume(ubi, vol->vol_id);
391 if (err < 0)
392 return err;
393
394 if (err) {
395 ubi_warn("volume %d on UBI device %d is corrupted",
396 vol->vol_id, ubi->ubi_num);
397 vol->corrupted = 1;
398 }
399 vol->checked = 1;
400 revoke_exclusive(desc, UBI_READWRITE);
401 }
402
403 *offp += count;
404 return count;
405}
406
407static int vol_cdev_ioctl(struct inode *inode, struct file *file,
408 unsigned int cmd, unsigned long arg)
409{
410 int err = 0;
411 struct ubi_volume_desc *desc = file->private_data;
412 struct ubi_volume *vol = desc->vol;
413 struct ubi_device *ubi = vol->ubi;
414 void __user *argp = (void __user *)arg;
415
416 if (_IOC_NR(cmd) > VOL_CDEV_IOC_MAX_SEQ ||
417 _IOC_TYPE(cmd) != UBI_VOL_IOC_MAGIC)
418 return -ENOTTY;
419
420 if (_IOC_DIR(cmd) && _IOC_READ)
421 err = !access_ok(VERIFY_WRITE, argp, _IOC_SIZE(cmd));
422 else if (_IOC_DIR(cmd) && _IOC_WRITE)
423 err = !access_ok(VERIFY_READ, argp, _IOC_SIZE(cmd));
424 if (err)
425 return -EFAULT;
426
427 switch (cmd) {
428
429 /* Volume update command */
430 case UBI_IOCVOLUP:
431 {
432 int64_t bytes, rsvd_bytes;
433
434 if (!capable(CAP_SYS_RESOURCE)) {
435 err = -EPERM;
436 break;
437 }
438
439 err = copy_from_user(&bytes, argp, sizeof(int64_t));
440 if (err) {
441 err = -EFAULT;
442 break;
443 }
444
445 if (desc->mode == UBI_READONLY) {
446 err = -EROFS;
447 break;
448 }
449
450 rsvd_bytes = vol->reserved_pebs * (ubi->leb_size-vol->data_pad);
451 if (bytes < 0 || bytes > rsvd_bytes) {
452 err = -EINVAL;
453 break;
454 }
455
456 err = get_exclusive(desc);
457 if (err < 0)
458 break;
459
460 err = ubi_start_update(ubi, vol->vol_id, bytes);
461 if (bytes == 0)
462 revoke_exclusive(desc, UBI_READWRITE);
463
464 file->f_pos = 0;
465 break;
466 }
467
468#ifdef CONFIG_MTD_UBI_DEBUG_USERSPACE_IO
469 /* Logical eraseblock erasure command */
470 case UBI_IOCEBER:
471 {
472 int32_t lnum;
473
474 err = __get_user(lnum, (__user int32_t *)argp);
475 if (err) {
476 err = -EFAULT;
477 break;
478 }
479
480 if (desc->mode == UBI_READONLY) {
481 err = -EROFS;
482 break;
483 }
484
485 if (lnum < 0 || lnum >= vol->reserved_pebs) {
486 err = -EINVAL;
487 break;
488 }
489
490 if (vol->vol_type != UBI_DYNAMIC_VOLUME) {
491 err = -EROFS;
492 break;
493 }
494
495 dbg_msg("erase LEB %d:%d", vol->vol_id, lnum);
496 err = ubi_eba_unmap_leb(ubi, vol->vol_id, lnum);
497 if (err)
498 break;
499
500 err = ubi_wl_flush(ubi);
501 break;
502 }
503#endif
504
505 default:
506 err = -ENOTTY;
507 break;
508 }
509
510 return err;
511}
512
513/**
514 * verify_mkvol_req - verify volume creation request.
515 * @ubi: UBI device description object
516 * @req: the request to check
517 *
518 * This function zero if the request is correct, and %-EINVAL if not.
519 */
520static int verify_mkvol_req(const struct ubi_device *ubi,
521 const struct ubi_mkvol_req *req)
522{
523 int n, err = -EINVAL;
524
525 if (req->bytes < 0 || req->alignment < 0 || req->vol_type < 0 ||
526 req->name_len < 0)
527 goto bad;
528
529 if ((req->vol_id < 0 || req->vol_id >= ubi->vtbl_slots) &&
530 req->vol_id != UBI_VOL_NUM_AUTO)
531 goto bad;
532
533 if (req->alignment == 0)
534 goto bad;
535
536 if (req->bytes == 0)
537 goto bad;
538
539 if (req->vol_type != UBI_DYNAMIC_VOLUME &&
540 req->vol_type != UBI_STATIC_VOLUME)
541 goto bad;
542
543 if (req->alignment > ubi->leb_size)
544 goto bad;
545
546 n = req->alignment % ubi->min_io_size;
547 if (req->alignment != 1 && n)
548 goto bad;
549
550 if (req->name_len > UBI_VOL_NAME_MAX) {
551 err = -ENAMETOOLONG;
552 goto bad;
553 }
554
555 return 0;
556
557bad:
558 dbg_err("bad volume creation request");
559 ubi_dbg_dump_mkvol_req(req);
560 return err;
561}
562
563/**
564 * verify_rsvol_req - verify volume re-size request.
565 * @ubi: UBI device description object
566 * @req: the request to check
567 *
568 * This function returns zero if the request is correct, and %-EINVAL if not.
569 */
570static int verify_rsvol_req(const struct ubi_device *ubi,
571 const struct ubi_rsvol_req *req)
572{
573 if (req->bytes <= 0)
574 return -EINVAL;
575
576 if (req->vol_id < 0 || req->vol_id >= ubi->vtbl_slots)
577 return -EINVAL;
578
579 return 0;
580}
581
582static int ubi_cdev_ioctl(struct inode *inode, struct file *file,
583 unsigned int cmd, unsigned long arg)
584{
585 int err = 0;
586 struct ubi_device *ubi;
587 struct ubi_volume_desc *desc;
588 void __user *argp = (void __user *)arg;
589
590 if (_IOC_NR(cmd) > UBI_CDEV_IOC_MAX_SEQ ||
591 _IOC_TYPE(cmd) != UBI_IOC_MAGIC)
592 return -ENOTTY;
593
594 if (_IOC_DIR(cmd) && _IOC_READ)
595 err = !access_ok(VERIFY_WRITE, argp, _IOC_SIZE(cmd));
596 else if (_IOC_DIR(cmd) && _IOC_WRITE)
597 err = !access_ok(VERIFY_READ, argp, _IOC_SIZE(cmd));
598 if (err)
599 return -EFAULT;
600
601 if (!capable(CAP_SYS_RESOURCE))
602 return -EPERM;
603
604 ubi = major_to_device(imajor(inode));
605 if (IS_ERR(ubi))
606 return PTR_ERR(ubi);
607
608 switch (cmd) {
609 /* Create volume command */
610 case UBI_IOCMKVOL:
611 {
612 struct ubi_mkvol_req req;
613
614 dbg_msg("create volume");
615 err = __copy_from_user(&req, argp,
616 sizeof(struct ubi_mkvol_req));
617 if (err) {
618 err = -EFAULT;
619 break;
620 }
621
622 err = verify_mkvol_req(ubi, &req);
623 if (err)
624 break;
625
626 req.name[req.name_len] = '\0';
627
628 err = ubi_create_volume(ubi, &req);
629 if (err)
630 break;
631
632 err = __put_user(req.vol_id, (__user int32_t *)argp);
633 if (err)
634 err = -EFAULT;
635
636 break;
637 }
638
639 /* Remove volume command */
640 case UBI_IOCRMVOL:
641 {
642 int vol_id;
643
644 dbg_msg("remove volume");
645 err = __get_user(vol_id, (__user int32_t *)argp);
646 if (err) {
647 err = -EFAULT;
648 break;
649 }
650
651 desc = ubi_open_volume(ubi->ubi_num, vol_id, UBI_EXCLUSIVE);
652 if (IS_ERR(desc)) {
653 err = PTR_ERR(desc);
654 break;
655 }
656
657 err = ubi_remove_volume(desc);
658 if (err)
659 ubi_close_volume(desc);
660
661 break;
662 }
663
664 /* Re-size volume command */
665 case UBI_IOCRSVOL:
666 {
667 int pebs;
668 uint64_t tmp;
669 struct ubi_rsvol_req req;
670
671 dbg_msg("re-size volume");
672 err = __copy_from_user(&req, argp,
673 sizeof(struct ubi_rsvol_req));
674 if (err) {
675 err = -EFAULT;
676 break;
677 }
678
679 err = verify_rsvol_req(ubi, &req);
680 if (err)
681 break;
682
683 desc = ubi_open_volume(ubi->ubi_num, req.vol_id, UBI_EXCLUSIVE);
684 if (IS_ERR(desc)) {
685 err = PTR_ERR(desc);
686 break;
687 }
688
689 tmp = req.bytes;
690 pebs = !!do_div(tmp, desc->vol->usable_leb_size);
691 pebs += tmp;
692
693 err = ubi_resize_volume(desc, pebs);
694 ubi_close_volume(desc);
695 break;
696 }
697
698 default:
699 err = -ENOTTY;
700 break;
701 }
702
703 return err;
704}
705
706/* UBI character device operations */
707struct file_operations ubi_cdev_operations = {
708 .owner = THIS_MODULE,
709 .ioctl = ubi_cdev_ioctl,
710 .llseek = no_llseek
711};
712
713/* UBI volume character device operations */
714struct file_operations ubi_vol_cdev_operations = {
715 .owner = THIS_MODULE,
716 .open = vol_cdev_open,
717 .release = vol_cdev_release,
718 .llseek = vol_cdev_llseek,
719 .read = vol_cdev_read,
720 .write = vol_cdev_write,
721 .ioctl = vol_cdev_ioctl
722};
diff --git a/drivers/mtd/ubi/debug.c b/drivers/mtd/ubi/debug.c
new file mode 100644
index 000000000000..86364221fafe
--- /dev/null
+++ b/drivers/mtd/ubi/debug.c
@@ -0,0 +1,224 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/*
22 * Here we keep all the UBI debugging stuff which should normally be disabled
23 * and compiled-out, but it is extremely helpful when hunting bugs or doing big
24 * changes.
25 */
26
27#ifdef CONFIG_MTD_UBI_DEBUG_MSG
28
29#include "ubi.h"
30
31/**
32 * ubi_dbg_dump_ec_hdr - dump an erase counter header.
33 * @ec_hdr: the erase counter header to dump
34 */
35void ubi_dbg_dump_ec_hdr(const struct ubi_ec_hdr *ec_hdr)
36{
37 dbg_msg("erase counter header dump:");
38 dbg_msg("magic %#08x", ubi32_to_cpu(ec_hdr->magic));
39 dbg_msg("version %d", (int)ec_hdr->version);
40 dbg_msg("ec %llu", (long long)ubi64_to_cpu(ec_hdr->ec));
41 dbg_msg("vid_hdr_offset %d", ubi32_to_cpu(ec_hdr->vid_hdr_offset));
42 dbg_msg("data_offset %d", ubi32_to_cpu(ec_hdr->data_offset));
43 dbg_msg("hdr_crc %#08x", ubi32_to_cpu(ec_hdr->hdr_crc));
44 dbg_msg("erase counter header hexdump:");
45 ubi_dbg_hexdump(ec_hdr, UBI_EC_HDR_SIZE);
46}
47
48/**
49 * ubi_dbg_dump_vid_hdr - dump a volume identifier header.
50 * @vid_hdr: the volume identifier header to dump
51 */
52void ubi_dbg_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr)
53{
54 dbg_msg("volume identifier header dump:");
55 dbg_msg("magic %08x", ubi32_to_cpu(vid_hdr->magic));
56 dbg_msg("version %d", (int)vid_hdr->version);
57 dbg_msg("vol_type %d", (int)vid_hdr->vol_type);
58 dbg_msg("copy_flag %d", (int)vid_hdr->copy_flag);
59 dbg_msg("compat %d", (int)vid_hdr->compat);
60 dbg_msg("vol_id %d", ubi32_to_cpu(vid_hdr->vol_id));
61 dbg_msg("lnum %d", ubi32_to_cpu(vid_hdr->lnum));
62 dbg_msg("leb_ver %u", ubi32_to_cpu(vid_hdr->leb_ver));
63 dbg_msg("data_size %d", ubi32_to_cpu(vid_hdr->data_size));
64 dbg_msg("used_ebs %d", ubi32_to_cpu(vid_hdr->used_ebs));
65 dbg_msg("data_pad %d", ubi32_to_cpu(vid_hdr->data_pad));
66 dbg_msg("sqnum %llu",
67 (unsigned long long)ubi64_to_cpu(vid_hdr->sqnum));
68 dbg_msg("hdr_crc %08x", ubi32_to_cpu(vid_hdr->hdr_crc));
69 dbg_msg("volume identifier header hexdump:");
70}
71
72/**
73 * ubi_dbg_dump_vol_info- dump volume information.
74 * @vol: UBI volume description object
75 */
76void ubi_dbg_dump_vol_info(const struct ubi_volume *vol)
77{
78 dbg_msg("volume information dump:");
79 dbg_msg("vol_id %d", vol->vol_id);
80 dbg_msg("reserved_pebs %d", vol->reserved_pebs);
81 dbg_msg("alignment %d", vol->alignment);
82 dbg_msg("data_pad %d", vol->data_pad);
83 dbg_msg("vol_type %d", vol->vol_type);
84 dbg_msg("name_len %d", vol->name_len);
85 dbg_msg("usable_leb_size %d", vol->usable_leb_size);
86 dbg_msg("used_ebs %d", vol->used_ebs);
87 dbg_msg("used_bytes %lld", vol->used_bytes);
88 dbg_msg("last_eb_bytes %d", vol->last_eb_bytes);
89 dbg_msg("corrupted %d", vol->corrupted);
90 dbg_msg("upd_marker %d", vol->upd_marker);
91
92 if (vol->name_len <= UBI_VOL_NAME_MAX &&
93 strnlen(vol->name, vol->name_len + 1) == vol->name_len) {
94 dbg_msg("name %s", vol->name);
95 } else {
96 dbg_msg("the 1st 5 characters of the name: %c%c%c%c%c",
97 vol->name[0], vol->name[1], vol->name[2],
98 vol->name[3], vol->name[4]);
99 }
100}
101
102/**
103 * ubi_dbg_dump_vtbl_record - dump a &struct ubi_vtbl_record object.
104 * @r: the object to dump
105 * @idx: volume table index
106 */
107void ubi_dbg_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx)
108{
109 int name_len = ubi16_to_cpu(r->name_len);
110
111 dbg_msg("volume table record %d dump:", idx);
112 dbg_msg("reserved_pebs %d", ubi32_to_cpu(r->reserved_pebs));
113 dbg_msg("alignment %d", ubi32_to_cpu(r->alignment));
114 dbg_msg("data_pad %d", ubi32_to_cpu(r->data_pad));
115 dbg_msg("vol_type %d", (int)r->vol_type);
116 dbg_msg("upd_marker %d", (int)r->upd_marker);
117 dbg_msg("name_len %d", name_len);
118
119 if (r->name[0] == '\0') {
120 dbg_msg("name NULL");
121 return;
122 }
123
124 if (name_len <= UBI_VOL_NAME_MAX &&
125 strnlen(&r->name[0], name_len + 1) == name_len) {
126 dbg_msg("name %s", &r->name[0]);
127 } else {
128 dbg_msg("1st 5 characters of the name: %c%c%c%c%c",
129 r->name[0], r->name[1], r->name[2], r->name[3],
130 r->name[4]);
131 }
132 dbg_msg("crc %#08x", ubi32_to_cpu(r->crc));
133}
134
135/**
136 * ubi_dbg_dump_sv - dump a &struct ubi_scan_volume object.
137 * @sv: the object to dump
138 */
139void ubi_dbg_dump_sv(const struct ubi_scan_volume *sv)
140{
141 dbg_msg("volume scanning information dump:");
142 dbg_msg("vol_id %d", sv->vol_id);
143 dbg_msg("highest_lnum %d", sv->highest_lnum);
144 dbg_msg("leb_count %d", sv->leb_count);
145 dbg_msg("compat %d", sv->compat);
146 dbg_msg("vol_type %d", sv->vol_type);
147 dbg_msg("used_ebs %d", sv->used_ebs);
148 dbg_msg("last_data_size %d", sv->last_data_size);
149 dbg_msg("data_pad %d", sv->data_pad);
150}
151
152/**
153 * ubi_dbg_dump_seb - dump a &struct ubi_scan_leb object.
154 * @seb: the object to dump
155 * @type: object type: 0 - not corrupted, 1 - corrupted
156 */
157void ubi_dbg_dump_seb(const struct ubi_scan_leb *seb, int type)
158{
159 dbg_msg("eraseblock scanning information dump:");
160 dbg_msg("ec %d", seb->ec);
161 dbg_msg("pnum %d", seb->pnum);
162 if (type == 0) {
163 dbg_msg("lnum %d", seb->lnum);
164 dbg_msg("scrub %d", seb->scrub);
165 dbg_msg("sqnum %llu", seb->sqnum);
166 dbg_msg("leb_ver %u", seb->leb_ver);
167 }
168}
169
170/**
171 * ubi_dbg_dump_mkvol_req - dump a &struct ubi_mkvol_req object.
172 * @req: the object to dump
173 */
174void ubi_dbg_dump_mkvol_req(const struct ubi_mkvol_req *req)
175{
176 char nm[17];
177
178 dbg_msg("volume creation request dump:");
179 dbg_msg("vol_id %d", req->vol_id);
180 dbg_msg("alignment %d", req->alignment);
181 dbg_msg("bytes %lld", (long long)req->bytes);
182 dbg_msg("vol_type %d", req->vol_type);
183 dbg_msg("name_len %d", req->name_len);
184
185 memcpy(nm, req->name, 16);
186 nm[16] = 0;
187 dbg_msg("the 1st 16 characters of the name: %s", nm);
188}
189
190#define BYTES_PER_LINE 32
191
192/**
193 * ubi_dbg_hexdump - dump a buffer.
194 * @ptr: the buffer to dump
195 * @size: buffer size which must be multiple of 4 bytes
196 */
197void ubi_dbg_hexdump(const void *ptr, int size)
198{
199 int i, k = 0, rows, columns;
200 const uint8_t *p = ptr;
201
202 size = ALIGN(size, 4);
203 rows = size/BYTES_PER_LINE + size % BYTES_PER_LINE;
204 for (i = 0; i < rows; i++) {
205 int j;
206
207 cond_resched();
208 columns = min(size - k, BYTES_PER_LINE) / 4;
209 if (columns == 0)
210 break;
211 printk(KERN_DEBUG "%5d: ", i * BYTES_PER_LINE);
212 for (j = 0; j < columns; j++) {
213 int n, N;
214
215 N = size - k > 4 ? 4 : size - k;
216 for (n = 0; n < N; n++)
217 printk("%02x", p[k++]);
218 printk(" ");
219 }
220 printk("\n");
221 }
222}
223
224#endif /* CONFIG_MTD_UBI_DEBUG_MSG */
diff --git a/drivers/mtd/ubi/debug.h b/drivers/mtd/ubi/debug.h
new file mode 100644
index 000000000000..f816ad9a36c0
--- /dev/null
+++ b/drivers/mtd/ubi/debug.h
@@ -0,0 +1,161 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21#ifndef __UBI_DEBUG_H__
22#define __UBI_DEBUG_H__
23
24#ifdef CONFIG_MTD_UBI_DEBUG
25#include <linux/random.h>
26
27#define ubi_assert(expr) BUG_ON(!(expr))
28#define dbg_err(fmt, ...) ubi_err(fmt, ##__VA_ARGS__)
29#else
30#define ubi_assert(expr) ({})
31#define dbg_err(fmt, ...) ({})
32#endif
33
34#ifdef CONFIG_MTD_UBI_DEBUG_DISABLE_BGT
35#define DBG_DISABLE_BGT 1
36#else
37#define DBG_DISABLE_BGT 0
38#endif
39
40#ifdef CONFIG_MTD_UBI_DEBUG_MSG
41/* Generic debugging message */
42#define dbg_msg(fmt, ...) \
43 printk(KERN_DEBUG "UBI DBG: %s: " fmt "\n", __FUNCTION__, ##__VA_ARGS__)
44
45#define ubi_dbg_dump_stack() dump_stack()
46
47struct ubi_ec_hdr;
48struct ubi_vid_hdr;
49struct ubi_volume;
50struct ubi_vtbl_record;
51struct ubi_scan_volume;
52struct ubi_scan_leb;
53struct ubi_mkvol_req;
54
55void ubi_dbg_print(int type, const char *func, const char *fmt, ...);
56void ubi_dbg_dump_ec_hdr(const struct ubi_ec_hdr *ec_hdr);
57void ubi_dbg_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr);
58void ubi_dbg_dump_vol_info(const struct ubi_volume *vol);
59void ubi_dbg_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx);
60void ubi_dbg_dump_sv(const struct ubi_scan_volume *sv);
61void ubi_dbg_dump_seb(const struct ubi_scan_leb *seb, int type);
62void ubi_dbg_dump_mkvol_req(const struct ubi_mkvol_req *req);
63void ubi_dbg_hexdump(const void *buf, int size);
64
65#else
66
67#define dbg_msg(fmt, ...) ({})
68#define ubi_dbg_dump_stack() ({})
69#define ubi_dbg_print(func, fmt, ...) ({})
70#define ubi_dbg_dump_ec_hdr(ec_hdr) ({})
71#define ubi_dbg_dump_vid_hdr(vid_hdr) ({})
72#define ubi_dbg_dump_vol_info(vol) ({})
73#define ubi_dbg_dump_vtbl_record(r, idx) ({})
74#define ubi_dbg_dump_sv(sv) ({})
75#define ubi_dbg_dump_seb(seb, type) ({})
76#define ubi_dbg_dump_mkvol_req(req) ({})
77#define ubi_dbg_hexdump(buf, size) ({})
78
79#endif /* CONFIG_MTD_UBI_DEBUG_MSG */
80
81#ifdef CONFIG_MTD_UBI_DEBUG_MSG_EBA
82/* Messages from the eraseblock association unit */
83#define dbg_eba(fmt, ...) \
84 printk(KERN_DEBUG "UBI DBG eba: %s: " fmt "\n", __FUNCTION__, \
85 ##__VA_ARGS__)
86#else
87#define dbg_eba(fmt, ...) ({})
88#endif
89
90#ifdef CONFIG_MTD_UBI_DEBUG_MSG_WL
91/* Messages from the wear-leveling unit */
92#define dbg_wl(fmt, ...) \
93 printk(KERN_DEBUG "UBI DBG wl: %s: " fmt "\n", __FUNCTION__, \
94 ##__VA_ARGS__)
95#else
96#define dbg_wl(fmt, ...) ({})
97#endif
98
99#ifdef CONFIG_MTD_UBI_DEBUG_MSG_IO
100/* Messages from the input/output unit */
101#define dbg_io(fmt, ...) \
102 printk(KERN_DEBUG "UBI DBG io: %s: " fmt "\n", __FUNCTION__, \
103 ##__VA_ARGS__)
104#else
105#define dbg_io(fmt, ...) ({})
106#endif
107
108#ifdef CONFIG_MTD_UBI_DEBUG_MSG_BLD
109/* Initialization and build messages */
110#define dbg_bld(fmt, ...) \
111 printk(KERN_DEBUG "UBI DBG bld: %s: " fmt "\n", __FUNCTION__, \
112 ##__VA_ARGS__)
113#else
114#define dbg_bld(fmt, ...) ({})
115#endif
116
117#ifdef CONFIG_MTD_UBI_DEBUG_EMULATE_BITFLIPS
118/**
119 * ubi_dbg_is_bitflip - if it is time to emulate a bit-flip.
120 *
121 * Returns non-zero if a bit-flip should be emulated, otherwise returns zero.
122 */
123static inline int ubi_dbg_is_bitflip(void)
124{
125 return !(random32() % 200);
126}
127#else
128#define ubi_dbg_is_bitflip() 0
129#endif
130
131#ifdef CONFIG_MTD_UBI_DEBUG_EMULATE_WRITE_FAILURES
132/**
133 * ubi_dbg_is_write_failure - if it is time to emulate a write failure.
134 *
135 * Returns non-zero if a write failure should be emulated, otherwise returns
136 * zero.
137 */
138static inline int ubi_dbg_is_write_failure(void)
139{
140 return !(random32() % 500);
141}
142#else
143#define ubi_dbg_is_write_failure() 0
144#endif
145
146#ifdef CONFIG_MTD_UBI_DEBUG_EMULATE_ERASE_FAILURES
147/**
148 * ubi_dbg_is_erase_failure - if its time to emulate an erase failure.
149 *
150 * Returns non-zero if an erase failure should be emulated, otherwise returns
151 * zero.
152 */
153static inline int ubi_dbg_is_erase_failure(void)
154{
155 return !(random32() % 400);
156}
157#else
158#define ubi_dbg_is_erase_failure() 0
159#endif
160
161#endif /* !__UBI_DEBUG_H__ */
diff --git a/drivers/mtd/ubi/eba.c b/drivers/mtd/ubi/eba.c
new file mode 100644
index 000000000000..d847ee1da3d9
--- /dev/null
+++ b/drivers/mtd/ubi/eba.c
@@ -0,0 +1,1241 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/*
22 * The UBI Eraseblock Association (EBA) unit.
23 *
24 * This unit is responsible for I/O to/from logical eraseblock.
25 *
26 * Although in this implementation the EBA table is fully kept and managed in
27 * RAM, which assumes poor scalability, it might be (partially) maintained on
28 * flash in future implementations.
29 *
30 * The EBA unit implements per-logical eraseblock locking. Before accessing a
31 * logical eraseblock it is locked for reading or writing. The per-logical
32 * eraseblock locking is implemented by means of the lock tree. The lock tree
33 * is an RB-tree which refers all the currently locked logical eraseblocks. The
34 * lock tree elements are &struct ltree_entry objects. They are indexed by
35 * (@vol_id, @lnum) pairs.
36 *
37 * EBA also maintains the global sequence counter which is incremented each
38 * time a logical eraseblock is mapped to a physical eraseblock and it is
39 * stored in the volume identifier header. This means that each VID header has
40 * a unique sequence number. The sequence number is only increased an we assume
41 * 64 bits is enough to never overflow.
42 */
43
44#include <linux/slab.h>
45#include <linux/crc32.h>
46#include <linux/err.h>
47#include "ubi.h"
48
49/**
50 * struct ltree_entry - an entry in the lock tree.
51 * @rb: links RB-tree nodes
52 * @vol_id: volume ID of the locked logical eraseblock
53 * @lnum: locked logical eraseblock number
54 * @users: how many tasks are using this logical eraseblock or wait for it
55 * @mutex: read/write mutex to implement read/write access serialization to
56 * the (@vol_id, @lnum) logical eraseblock
57 *
58 * When a logical eraseblock is being locked - corresponding &struct ltree_entry
59 * object is inserted to the lock tree (@ubi->ltree).
60 */
61struct ltree_entry {
62 struct rb_node rb;
63 int vol_id;
64 int lnum;
65 int users;
66 struct rw_semaphore mutex;
67};
68
69/* Slab cache for lock-tree entries */
70static struct kmem_cache *ltree_slab;
71
72/**
73 * next_sqnum - get next sequence number.
74 * @ubi: UBI device description object
75 *
76 * This function returns next sequence number to use, which is just the current
77 * global sequence counter value. It also increases the global sequence
78 * counter.
79 */
80static unsigned long long next_sqnum(struct ubi_device *ubi)
81{
82 unsigned long long sqnum;
83
84 spin_lock(&ubi->ltree_lock);
85 sqnum = ubi->global_sqnum++;
86 spin_unlock(&ubi->ltree_lock);
87
88 return sqnum;
89}
90
91/**
92 * ubi_get_compat - get compatibility flags of a volume.
93 * @ubi: UBI device description object
94 * @vol_id: volume ID
95 *
96 * This function returns compatibility flags for an internal volume. User
97 * volumes have no compatibility flags, so %0 is returned.
98 */
99static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
100{
101 if (vol_id == UBI_LAYOUT_VOL_ID)
102 return UBI_LAYOUT_VOLUME_COMPAT;
103 return 0;
104}
105
106/**
107 * ltree_lookup - look up the lock tree.
108 * @ubi: UBI device description object
109 * @vol_id: volume ID
110 * @lnum: logical eraseblock number
111 *
112 * This function returns a pointer to the corresponding &struct ltree_entry
113 * object if the logical eraseblock is locked and %NULL if it is not.
114 * @ubi->ltree_lock has to be locked.
115 */
116static struct ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
117 int lnum)
118{
119 struct rb_node *p;
120
121 p = ubi->ltree.rb_node;
122 while (p) {
123 struct ltree_entry *le;
124
125 le = rb_entry(p, struct ltree_entry, rb);
126
127 if (vol_id < le->vol_id)
128 p = p->rb_left;
129 else if (vol_id > le->vol_id)
130 p = p->rb_right;
131 else {
132 if (lnum < le->lnum)
133 p = p->rb_left;
134 else if (lnum > le->lnum)
135 p = p->rb_right;
136 else
137 return le;
138 }
139 }
140
141 return NULL;
142}
143
144/**
145 * ltree_add_entry - add new entry to the lock tree.
146 * @ubi: UBI device description object
147 * @vol_id: volume ID
148 * @lnum: logical eraseblock number
149 *
150 * This function adds new entry for logical eraseblock (@vol_id, @lnum) to the
151 * lock tree. If such entry is already there, its usage counter is increased.
152 * Returns pointer to the lock tree entry or %-ENOMEM if memory allocation
153 * failed.
154 */
155static struct ltree_entry *ltree_add_entry(struct ubi_device *ubi, int vol_id,
156 int lnum)
157{
158 struct ltree_entry *le, *le1, *le_free;
159
160 le = kmem_cache_alloc(ltree_slab, GFP_KERNEL);
161 if (!le)
162 return ERR_PTR(-ENOMEM);
163
164 le->vol_id = vol_id;
165 le->lnum = lnum;
166
167 spin_lock(&ubi->ltree_lock);
168 le1 = ltree_lookup(ubi, vol_id, lnum);
169
170 if (le1) {
171 /*
172 * This logical eraseblock is already locked. The newly
173 * allocated lock entry is not needed.
174 */
175 le_free = le;
176 le = le1;
177 } else {
178 struct rb_node **p, *parent = NULL;
179
180 /*
181 * No lock entry, add the newly allocated one to the
182 * @ubi->ltree RB-tree.
183 */
184 le_free = NULL;
185
186 p = &ubi->ltree.rb_node;
187 while (*p) {
188 parent = *p;
189 le1 = rb_entry(parent, struct ltree_entry, rb);
190
191 if (vol_id < le1->vol_id)
192 p = &(*p)->rb_left;
193 else if (vol_id > le1->vol_id)
194 p = &(*p)->rb_right;
195 else {
196 ubi_assert(lnum != le1->lnum);
197 if (lnum < le1->lnum)
198 p = &(*p)->rb_left;
199 else
200 p = &(*p)->rb_right;
201 }
202 }
203
204 rb_link_node(&le->rb, parent, p);
205 rb_insert_color(&le->rb, &ubi->ltree);
206 }
207 le->users += 1;
208 spin_unlock(&ubi->ltree_lock);
209
210 if (le_free)
211 kmem_cache_free(ltree_slab, le_free);
212
213 return le;
214}
215
216/**
217 * leb_read_lock - lock logical eraseblock for reading.
218 * @ubi: UBI device description object
219 * @vol_id: volume ID
220 * @lnum: logical eraseblock number
221 *
222 * This function locks a logical eraseblock for reading. Returns zero in case
223 * of success and a negative error code in case of failure.
224 */
225static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
226{
227 struct ltree_entry *le;
228
229 le = ltree_add_entry(ubi, vol_id, lnum);
230 if (IS_ERR(le))
231 return PTR_ERR(le);
232 down_read(&le->mutex);
233 return 0;
234}
235
236/**
237 * leb_read_unlock - unlock logical eraseblock.
238 * @ubi: UBI device description object
239 * @vol_id: volume ID
240 * @lnum: logical eraseblock number
241 */
242static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
243{
244 int free = 0;
245 struct ltree_entry *le;
246
247 spin_lock(&ubi->ltree_lock);
248 le = ltree_lookup(ubi, vol_id, lnum);
249 le->users -= 1;
250 ubi_assert(le->users >= 0);
251 if (le->users == 0) {
252 rb_erase(&le->rb, &ubi->ltree);
253 free = 1;
254 }
255 spin_unlock(&ubi->ltree_lock);
256
257 up_read(&le->mutex);
258 if (free)
259 kmem_cache_free(ltree_slab, le);
260}
261
262/**
263 * leb_write_lock - lock logical eraseblock for writing.
264 * @ubi: UBI device description object
265 * @vol_id: volume ID
266 * @lnum: logical eraseblock number
267 *
268 * This function locks a logical eraseblock for writing. Returns zero in case
269 * of success and a negative error code in case of failure.
270 */
271static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
272{
273 struct ltree_entry *le;
274
275 le = ltree_add_entry(ubi, vol_id, lnum);
276 if (IS_ERR(le))
277 return PTR_ERR(le);
278 down_write(&le->mutex);
279 return 0;
280}
281
282/**
283 * leb_write_unlock - unlock logical eraseblock.
284 * @ubi: UBI device description object
285 * @vol_id: volume ID
286 * @lnum: logical eraseblock number
287 */
288static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
289{
290 int free;
291 struct ltree_entry *le;
292
293 spin_lock(&ubi->ltree_lock);
294 le = ltree_lookup(ubi, vol_id, lnum);
295 le->users -= 1;
296 ubi_assert(le->users >= 0);
297 if (le->users == 0) {
298 rb_erase(&le->rb, &ubi->ltree);
299 free = 1;
300 } else
301 free = 0;
302 spin_unlock(&ubi->ltree_lock);
303
304 up_write(&le->mutex);
305 if (free)
306 kmem_cache_free(ltree_slab, le);
307}
308
309/**
310 * ubi_eba_unmap_leb - un-map logical eraseblock.
311 * @ubi: UBI device description object
312 * @vol_id: volume ID
313 * @lnum: logical eraseblock number
314 *
315 * This function un-maps logical eraseblock @lnum and schedules corresponding
316 * physical eraseblock for erasure. Returns zero in case of success and a
317 * negative error code in case of failure.
318 */
319int ubi_eba_unmap_leb(struct ubi_device *ubi, int vol_id, int lnum)
320{
321 int idx = vol_id2idx(ubi, vol_id), err, pnum;
322 struct ubi_volume *vol = ubi->volumes[idx];
323
324 if (ubi->ro_mode)
325 return -EROFS;
326
327 err = leb_write_lock(ubi, vol_id, lnum);
328 if (err)
329 return err;
330
331 pnum = vol->eba_tbl[lnum];
332 if (pnum < 0)
333 /* This logical eraseblock is already unmapped */
334 goto out_unlock;
335
336 dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
337
338 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
339 err = ubi_wl_put_peb(ubi, pnum, 0);
340
341out_unlock:
342 leb_write_unlock(ubi, vol_id, lnum);
343 return err;
344}
345
346/**
347 * ubi_eba_read_leb - read data.
348 * @ubi: UBI device description object
349 * @vol_id: volume ID
350 * @lnum: logical eraseblock number
351 * @buf: buffer to store the read data
352 * @offset: offset from where to read
353 * @len: how many bytes to read
354 * @check: data CRC check flag
355 *
356 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
357 * bytes. The @check flag only makes sense for static volumes and forces
358 * eraseblock data CRC checking.
359 *
360 * In case of success this function returns zero. In case of a static volume,
361 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
362 * returned for any volume type if an ECC error was detected by the MTD device
363 * driver. Other negative error cored may be returned in case of other errors.
364 */
365int ubi_eba_read_leb(struct ubi_device *ubi, int vol_id, int lnum, void *buf,
366 int offset, int len, int check)
367{
368 int err, pnum, scrub = 0, idx = vol_id2idx(ubi, vol_id);
369 struct ubi_vid_hdr *vid_hdr;
370 struct ubi_volume *vol = ubi->volumes[idx];
371 uint32_t crc, crc1;
372
373 err = leb_read_lock(ubi, vol_id, lnum);
374 if (err)
375 return err;
376
377 pnum = vol->eba_tbl[lnum];
378 if (pnum < 0) {
379 /*
380 * The logical eraseblock is not mapped, fill the whole buffer
381 * with 0xFF bytes. The exception is static volumes for which
382 * it is an error to read unmapped logical eraseblocks.
383 */
384 dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
385 len, offset, vol_id, lnum);
386 leb_read_unlock(ubi, vol_id, lnum);
387 ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
388 memset(buf, 0xFF, len);
389 return 0;
390 }
391
392 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
393 len, offset, vol_id, lnum, pnum);
394
395 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
396 check = 0;
397
398retry:
399 if (check) {
400 vid_hdr = ubi_zalloc_vid_hdr(ubi);
401 if (!vid_hdr) {
402 err = -ENOMEM;
403 goto out_unlock;
404 }
405
406 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
407 if (err && err != UBI_IO_BITFLIPS) {
408 if (err > 0) {
409 /*
410 * The header is either absent or corrupted.
411 * The former case means there is a bug -
412 * switch to read-only mode just in case.
413 * The latter case means a real corruption - we
414 * may try to recover data. FIXME: but this is
415 * not implemented.
416 */
417 if (err == UBI_IO_BAD_VID_HDR) {
418 ubi_warn("bad VID header at PEB %d, LEB"
419 "%d:%d", pnum, vol_id, lnum);
420 err = -EBADMSG;
421 } else
422 ubi_ro_mode(ubi);
423 }
424 goto out_free;
425 } else if (err == UBI_IO_BITFLIPS)
426 scrub = 1;
427
428 ubi_assert(lnum < ubi32_to_cpu(vid_hdr->used_ebs));
429 ubi_assert(len == ubi32_to_cpu(vid_hdr->data_size));
430
431 crc = ubi32_to_cpu(vid_hdr->data_crc);
432 ubi_free_vid_hdr(ubi, vid_hdr);
433 }
434
435 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
436 if (err) {
437 if (err == UBI_IO_BITFLIPS) {
438 scrub = 1;
439 err = 0;
440 } else if (err == -EBADMSG) {
441 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
442 goto out_unlock;
443 scrub = 1;
444 if (!check) {
445 ubi_msg("force data checking");
446 check = 1;
447 goto retry;
448 }
449 } else
450 goto out_unlock;
451 }
452
453 if (check) {
454 crc1 = crc32(UBI_CRC32_INIT, buf, len);
455 if (crc1 != crc) {
456 ubi_warn("CRC error: calculated %#08x, must be %#08x",
457 crc1, crc);
458 err = -EBADMSG;
459 goto out_unlock;
460 }
461 }
462
463 if (scrub)
464 err = ubi_wl_scrub_peb(ubi, pnum);
465
466 leb_read_unlock(ubi, vol_id, lnum);
467 return err;
468
469out_free:
470 ubi_free_vid_hdr(ubi, vid_hdr);
471out_unlock:
472 leb_read_unlock(ubi, vol_id, lnum);
473 return err;
474}
475
476/**
477 * recover_peb - recover from write failure.
478 * @ubi: UBI device description object
479 * @pnum: the physical eraseblock to recover
480 * @vol_id: volume ID
481 * @lnum: logical eraseblock number
482 * @buf: data which was not written because of the write failure
483 * @offset: offset of the failed write
484 * @len: how many bytes should have been written
485 *
486 * This function is called in case of a write failure and moves all good data
487 * from the potentially bad physical eraseblock to a good physical eraseblock.
488 * This function also writes the data which was not written due to the failure.
489 * Returns new physical eraseblock number in case of success, and a negative
490 * error code in case of failure.
491 */
492static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
493 const void *buf, int offset, int len)
494{
495 int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
496 struct ubi_volume *vol = ubi->volumes[idx];
497 struct ubi_vid_hdr *vid_hdr;
498 unsigned char *new_buf;
499
500 vid_hdr = ubi_zalloc_vid_hdr(ubi);
501 if (!vid_hdr) {
502 return -ENOMEM;
503 }
504
505retry:
506 new_pnum = ubi_wl_get_peb(ubi, UBI_UNKNOWN);
507 if (new_pnum < 0) {
508 ubi_free_vid_hdr(ubi, vid_hdr);
509 return new_pnum;
510 }
511
512 ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum);
513
514 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
515 if (err && err != UBI_IO_BITFLIPS) {
516 if (err > 0)
517 err = -EIO;
518 goto out_put;
519 }
520
521 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
522 err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
523 if (err)
524 goto write_error;
525
526 data_size = offset + len;
527 new_buf = kmalloc(data_size, GFP_KERNEL);
528 if (!new_buf) {
529 err = -ENOMEM;
530 goto out_put;
531 }
532 memset(new_buf + offset, 0xFF, len);
533
534 /* Read everything before the area where the write failure happened */
535 if (offset > 0) {
536 err = ubi_io_read_data(ubi, new_buf, pnum, 0, offset);
537 if (err && err != UBI_IO_BITFLIPS) {
538 kfree(new_buf);
539 goto out_put;
540 }
541 }
542
543 memcpy(new_buf + offset, buf, len);
544
545 err = ubi_io_write_data(ubi, new_buf, new_pnum, 0, data_size);
546 if (err) {
547 kfree(new_buf);
548 goto write_error;
549 }
550
551 kfree(new_buf);
552 ubi_free_vid_hdr(ubi, vid_hdr);
553
554 vol->eba_tbl[lnum] = new_pnum;
555 ubi_wl_put_peb(ubi, pnum, 1);
556
557 ubi_msg("data was successfully recovered");
558 return 0;
559
560out_put:
561 ubi_wl_put_peb(ubi, new_pnum, 1);
562 ubi_free_vid_hdr(ubi, vid_hdr);
563 return err;
564
565write_error:
566 /*
567 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
568 * get another one.
569 */
570 ubi_warn("failed to write to PEB %d", new_pnum);
571 ubi_wl_put_peb(ubi, new_pnum, 1);
572 if (++tries > UBI_IO_RETRIES) {
573 ubi_free_vid_hdr(ubi, vid_hdr);
574 return err;
575 }
576 ubi_msg("try again");
577 goto retry;
578}
579
580/**
581 * ubi_eba_write_leb - write data to dynamic volume.
582 * @ubi: UBI device description object
583 * @vol_id: volume ID
584 * @lnum: logical eraseblock number
585 * @buf: the data to write
586 * @offset: offset within the logical eraseblock where to write
587 * @len: how many bytes to write
588 * @dtype: data type
589 *
590 * This function writes data to logical eraseblock @lnum of a dynamic volume
591 * @vol_id. Returns zero in case of success and a negative error code in case
592 * of failure. In case of error, it is possible that something was still
593 * written to the flash media, but may be some garbage.
594 */
595int ubi_eba_write_leb(struct ubi_device *ubi, int vol_id, int lnum,
596 const void *buf, int offset, int len, int dtype)
597{
598 int idx = vol_id2idx(ubi, vol_id), err, pnum, tries = 0;
599 struct ubi_volume *vol = ubi->volumes[idx];
600 struct ubi_vid_hdr *vid_hdr;
601
602 if (ubi->ro_mode)
603 return -EROFS;
604
605 err = leb_write_lock(ubi, vol_id, lnum);
606 if (err)
607 return err;
608
609 pnum = vol->eba_tbl[lnum];
610 if (pnum >= 0) {
611 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
612 len, offset, vol_id, lnum, pnum);
613
614 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
615 if (err) {
616 ubi_warn("failed to write data to PEB %d", pnum);
617 if (err == -EIO && ubi->bad_allowed)
618 err = recover_peb(ubi, pnum, vol_id, lnum, buf, offset, len);
619 if (err)
620 ubi_ro_mode(ubi);
621 }
622 leb_write_unlock(ubi, vol_id, lnum);
623 return err;
624 }
625
626 /*
627 * The logical eraseblock is not mapped. We have to get a free physical
628 * eraseblock and write the volume identifier header there first.
629 */
630 vid_hdr = ubi_zalloc_vid_hdr(ubi);
631 if (!vid_hdr) {
632 leb_write_unlock(ubi, vol_id, lnum);
633 return -ENOMEM;
634 }
635
636 vid_hdr->vol_type = UBI_VID_DYNAMIC;
637 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
638 vid_hdr->vol_id = cpu_to_ubi32(vol_id);
639 vid_hdr->lnum = cpu_to_ubi32(lnum);
640 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
641 vid_hdr->data_pad = cpu_to_ubi32(vol->data_pad);
642
643retry:
644 pnum = ubi_wl_get_peb(ubi, dtype);
645 if (pnum < 0) {
646 ubi_free_vid_hdr(ubi, vid_hdr);
647 leb_write_unlock(ubi, vol_id, lnum);
648 return pnum;
649 }
650
651 dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
652 len, offset, vol_id, lnum, pnum);
653
654 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
655 if (err) {
656 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
657 vol_id, lnum, pnum);
658 goto write_error;
659 }
660
661 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
662 if (err) {
663 ubi_warn("failed to write %d bytes at offset %d of LEB %d:%d, "
664 "PEB %d", len, offset, vol_id, lnum, pnum);
665 goto write_error;
666 }
667
668 vol->eba_tbl[lnum] = pnum;
669
670 leb_write_unlock(ubi, vol_id, lnum);
671 ubi_free_vid_hdr(ubi, vid_hdr);
672 return 0;
673
674write_error:
675 if (err != -EIO || !ubi->bad_allowed) {
676 ubi_ro_mode(ubi);
677 leb_write_unlock(ubi, vol_id, lnum);
678 ubi_free_vid_hdr(ubi, vid_hdr);
679 return err;
680 }
681
682 /*
683 * Fortunately, this is the first write operation to this physical
684 * eraseblock, so just put it and request a new one. We assume that if
685 * this physical eraseblock went bad, the erase code will handle that.
686 */
687 err = ubi_wl_put_peb(ubi, pnum, 1);
688 if (err || ++tries > UBI_IO_RETRIES) {
689 ubi_ro_mode(ubi);
690 leb_write_unlock(ubi, vol_id, lnum);
691 ubi_free_vid_hdr(ubi, vid_hdr);
692 return err;
693 }
694
695 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
696 ubi_msg("try another PEB");
697 goto retry;
698}
699
700/**
701 * ubi_eba_write_leb_st - write data to static volume.
702 * @ubi: UBI device description object
703 * @vol_id: volume ID
704 * @lnum: logical eraseblock number
705 * @buf: data to write
706 * @len: how many bytes to write
707 * @dtype: data type
708 * @used_ebs: how many logical eraseblocks will this volume contain
709 *
710 * This function writes data to logical eraseblock @lnum of static volume
711 * @vol_id. The @used_ebs argument should contain total number of logical
712 * eraseblock in this static volume.
713 *
714 * When writing to the last logical eraseblock, the @len argument doesn't have
715 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
716 * to the real data size, although the @buf buffer has to contain the
717 * alignment. In all other cases, @len has to be aligned.
718 *
719 * It is prohibited to write more then once to logical eraseblocks of static
720 * volumes. This function returns zero in case of success and a negative error
721 * code in case of failure.
722 */
723int ubi_eba_write_leb_st(struct ubi_device *ubi, int vol_id, int lnum,
724 const void *buf, int len, int dtype, int used_ebs)
725{
726 int err, pnum, tries = 0, data_size = len;
727 int idx = vol_id2idx(ubi, vol_id);
728 struct ubi_volume *vol = ubi->volumes[idx];
729 struct ubi_vid_hdr *vid_hdr;
730 uint32_t crc;
731
732 if (ubi->ro_mode)
733 return -EROFS;
734
735 if (lnum == used_ebs - 1)
736 /* If this is the last LEB @len may be unaligned */
737 len = ALIGN(data_size, ubi->min_io_size);
738 else
739 ubi_assert(len % ubi->min_io_size == 0);
740
741 vid_hdr = ubi_zalloc_vid_hdr(ubi);
742 if (!vid_hdr)
743 return -ENOMEM;
744
745 err = leb_write_lock(ubi, vol_id, lnum);
746 if (err) {
747 ubi_free_vid_hdr(ubi, vid_hdr);
748 return err;
749 }
750
751 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
752 vid_hdr->vol_id = cpu_to_ubi32(vol_id);
753 vid_hdr->lnum = cpu_to_ubi32(lnum);
754 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
755 vid_hdr->data_pad = cpu_to_ubi32(vol->data_pad);
756
757 crc = crc32(UBI_CRC32_INIT, buf, data_size);
758 vid_hdr->vol_type = UBI_VID_STATIC;
759 vid_hdr->data_size = cpu_to_ubi32(data_size);
760 vid_hdr->used_ebs = cpu_to_ubi32(used_ebs);
761 vid_hdr->data_crc = cpu_to_ubi32(crc);
762
763retry:
764 pnum = ubi_wl_get_peb(ubi, dtype);
765 if (pnum < 0) {
766 ubi_free_vid_hdr(ubi, vid_hdr);
767 leb_write_unlock(ubi, vol_id, lnum);
768 return pnum;
769 }
770
771 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
772 len, vol_id, lnum, pnum, used_ebs);
773
774 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
775 if (err) {
776 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
777 vol_id, lnum, pnum);
778 goto write_error;
779 }
780
781 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
782 if (err) {
783 ubi_warn("failed to write %d bytes of data to PEB %d",
784 len, pnum);
785 goto write_error;
786 }
787
788 ubi_assert(vol->eba_tbl[lnum] < 0);
789 vol->eba_tbl[lnum] = pnum;
790
791 leb_write_unlock(ubi, vol_id, lnum);
792 ubi_free_vid_hdr(ubi, vid_hdr);
793 return 0;
794
795write_error:
796 if (err != -EIO || !ubi->bad_allowed) {
797 /*
798 * This flash device does not admit of bad eraseblocks or
799 * something nasty and unexpected happened. Switch to read-only
800 * mode just in case.
801 */
802 ubi_ro_mode(ubi);
803 leb_write_unlock(ubi, vol_id, lnum);
804 ubi_free_vid_hdr(ubi, vid_hdr);
805 return err;
806 }
807
808 err = ubi_wl_put_peb(ubi, pnum, 1);
809 if (err || ++tries > UBI_IO_RETRIES) {
810 ubi_ro_mode(ubi);
811 leb_write_unlock(ubi, vol_id, lnum);
812 ubi_free_vid_hdr(ubi, vid_hdr);
813 return err;
814 }
815
816 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
817 ubi_msg("try another PEB");
818 goto retry;
819}
820
821/*
822 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
823 * @ubi: UBI device description object
824 * @vol_id: volume ID
825 * @lnum: logical eraseblock number
826 * @buf: data to write
827 * @len: how many bytes to write
828 * @dtype: data type
829 *
830 * This function changes the contents of a logical eraseblock atomically. @buf
831 * has to contain new logical eraseblock data, and @len - the length of the
832 * data, which has to be aligned. This function guarantees that in case of an
833 * unclean reboot the old contents is preserved. Returns zero in case of
834 * success and a negative error code in case of failure.
835 */
836int ubi_eba_atomic_leb_change(struct ubi_device *ubi, int vol_id, int lnum,
837 const void *buf, int len, int dtype)
838{
839 int err, pnum, tries = 0, idx = vol_id2idx(ubi, vol_id);
840 struct ubi_volume *vol = ubi->volumes[idx];
841 struct ubi_vid_hdr *vid_hdr;
842 uint32_t crc;
843
844 if (ubi->ro_mode)
845 return -EROFS;
846
847 vid_hdr = ubi_zalloc_vid_hdr(ubi);
848 if (!vid_hdr)
849 return -ENOMEM;
850
851 err = leb_write_lock(ubi, vol_id, lnum);
852 if (err) {
853 ubi_free_vid_hdr(ubi, vid_hdr);
854 return err;
855 }
856
857 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
858 vid_hdr->vol_id = cpu_to_ubi32(vol_id);
859 vid_hdr->lnum = cpu_to_ubi32(lnum);
860 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
861 vid_hdr->data_pad = cpu_to_ubi32(vol->data_pad);
862
863 crc = crc32(UBI_CRC32_INIT, buf, len);
864 vid_hdr->vol_type = UBI_VID_STATIC;
865 vid_hdr->data_size = cpu_to_ubi32(len);
866 vid_hdr->copy_flag = 1;
867 vid_hdr->data_crc = cpu_to_ubi32(crc);
868
869retry:
870 pnum = ubi_wl_get_peb(ubi, dtype);
871 if (pnum < 0) {
872 ubi_free_vid_hdr(ubi, vid_hdr);
873 leb_write_unlock(ubi, vol_id, lnum);
874 return pnum;
875 }
876
877 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
878 vol_id, lnum, vol->eba_tbl[lnum], pnum);
879
880 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
881 if (err) {
882 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
883 vol_id, lnum, pnum);
884 goto write_error;
885 }
886
887 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
888 if (err) {
889 ubi_warn("failed to write %d bytes of data to PEB %d",
890 len, pnum);
891 goto write_error;
892 }
893
894 err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 1);
895 if (err) {
896 ubi_free_vid_hdr(ubi, vid_hdr);
897 leb_write_unlock(ubi, vol_id, lnum);
898 return err;
899 }
900
901 vol->eba_tbl[lnum] = pnum;
902 leb_write_unlock(ubi, vol_id, lnum);
903 ubi_free_vid_hdr(ubi, vid_hdr);
904 return 0;
905
906write_error:
907 if (err != -EIO || !ubi->bad_allowed) {
908 /*
909 * This flash device does not admit of bad eraseblocks or
910 * something nasty and unexpected happened. Switch to read-only
911 * mode just in case.
912 */
913 ubi_ro_mode(ubi);
914 leb_write_unlock(ubi, vol_id, lnum);
915 ubi_free_vid_hdr(ubi, vid_hdr);
916 return err;
917 }
918
919 err = ubi_wl_put_peb(ubi, pnum, 1);
920 if (err || ++tries > UBI_IO_RETRIES) {
921 ubi_ro_mode(ubi);
922 leb_write_unlock(ubi, vol_id, lnum);
923 ubi_free_vid_hdr(ubi, vid_hdr);
924 return err;
925 }
926
927 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
928 ubi_msg("try another PEB");
929 goto retry;
930}
931
932/**
933 * ltree_entry_ctor - lock tree entries slab cache constructor.
934 * @obj: the lock-tree entry to construct
935 * @cache: the lock tree entry slab cache
936 * @flags: constructor flags
937 */
938static void ltree_entry_ctor(void *obj, struct kmem_cache *cache,
939 unsigned long flags)
940{
941 struct ltree_entry *le = obj;
942
943 if ((flags & (SLAB_CTOR_VERIFY | SLAB_CTOR_CONSTRUCTOR)) !=
944 SLAB_CTOR_CONSTRUCTOR)
945 return;
946
947 le->users = 0;
948 init_rwsem(&le->mutex);
949}
950
951/**
952 * ubi_eba_copy_leb - copy logical eraseblock.
953 * @ubi: UBI device description object
954 * @from: physical eraseblock number from where to copy
955 * @to: physical eraseblock number where to copy
956 * @vid_hdr: VID header of the @from physical eraseblock
957 *
958 * This function copies logical eraseblock from physical eraseblock @from to
959 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
960 * function. Returns zero in case of success, %UBI_IO_BITFLIPS if the operation
961 * was canceled because bit-flips were detected at the target PEB, and a
962 * negative error code in case of failure.
963 */
964int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
965 struct ubi_vid_hdr *vid_hdr)
966{
967 int err, vol_id, lnum, data_size, aldata_size, pnum, idx;
968 struct ubi_volume *vol;
969 uint32_t crc;
970 void *buf, *buf1 = NULL;
971
972 vol_id = ubi32_to_cpu(vid_hdr->vol_id);
973 lnum = ubi32_to_cpu(vid_hdr->lnum);
974
975 dbg_eba("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
976
977 if (vid_hdr->vol_type == UBI_VID_STATIC) {
978 data_size = ubi32_to_cpu(vid_hdr->data_size);
979 aldata_size = ALIGN(data_size, ubi->min_io_size);
980 } else
981 data_size = aldata_size =
982 ubi->leb_size - ubi32_to_cpu(vid_hdr->data_pad);
983
984 buf = kmalloc(aldata_size, GFP_KERNEL);
985 if (!buf)
986 return -ENOMEM;
987
988 /*
989 * We do not want anybody to write to this logical eraseblock while we
990 * are moving it, so we lock it.
991 */
992 err = leb_write_lock(ubi, vol_id, lnum);
993 if (err) {
994 kfree(buf);
995 return err;
996 }
997
998 /*
999 * But the logical eraseblock might have been put by this time.
1000 * Cancel if it is true.
1001 */
1002 idx = vol_id2idx(ubi, vol_id);
1003
1004 /*
1005 * We may race with volume deletion/re-size, so we have to hold
1006 * @ubi->volumes_lock.
1007 */
1008 spin_lock(&ubi->volumes_lock);
1009 vol = ubi->volumes[idx];
1010 if (!vol) {
1011 dbg_eba("volume %d was removed meanwhile", vol_id);
1012 spin_unlock(&ubi->volumes_lock);
1013 goto out_unlock;
1014 }
1015
1016 pnum = vol->eba_tbl[lnum];
1017 if (pnum != from) {
1018 dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1019 "PEB %d, cancel", vol_id, lnum, from, pnum);
1020 spin_unlock(&ubi->volumes_lock);
1021 goto out_unlock;
1022 }
1023 spin_unlock(&ubi->volumes_lock);
1024
1025 /* OK, now the LEB is locked and we can safely start moving it */
1026
1027 dbg_eba("read %d bytes of data", aldata_size);
1028 err = ubi_io_read_data(ubi, buf, from, 0, aldata_size);
1029 if (err && err != UBI_IO_BITFLIPS) {
1030 ubi_warn("error %d while reading data from PEB %d",
1031 err, from);
1032 goto out_unlock;
1033 }
1034
1035 /*
1036 * Now we have got to calculate how much data we have to to copy. In
1037 * case of a static volume it is fairly easy - the VID header contains
1038 * the data size. In case of a dynamic volume it is more difficult - we
1039 * have to read the contents, cut 0xFF bytes from the end and copy only
1040 * the first part. We must do this to avoid writing 0xFF bytes as it
1041 * may have some side-effects. And not only this. It is important not
1042 * to include those 0xFFs to CRC because later the they may be filled
1043 * by data.
1044 */
1045 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1046 aldata_size = data_size =
1047 ubi_calc_data_len(ubi, buf, data_size);
1048
1049 cond_resched();
1050 crc = crc32(UBI_CRC32_INIT, buf, data_size);
1051 cond_resched();
1052
1053 /*
1054 * It may turn out to me that the whole @from physical eraseblock
1055 * contains only 0xFF bytes. Then we have to only write the VID header
1056 * and do not write any data. This also means we should not set
1057 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1058 */
1059 if (data_size > 0) {
1060 vid_hdr->copy_flag = 1;
1061 vid_hdr->data_size = cpu_to_ubi32(data_size);
1062 vid_hdr->data_crc = cpu_to_ubi32(crc);
1063 }
1064 vid_hdr->sqnum = cpu_to_ubi64(next_sqnum(ubi));
1065
1066 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
1067 if (err)
1068 goto out_unlock;
1069
1070 cond_resched();
1071
1072 /* Read the VID header back and check if it was written correctly */
1073 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1074 if (err) {
1075 if (err != UBI_IO_BITFLIPS)
1076 ubi_warn("cannot read VID header back from PEB %d", to);
1077 goto out_unlock;
1078 }
1079
1080 if (data_size > 0) {
1081 err = ubi_io_write_data(ubi, buf, to, 0, aldata_size);
1082 if (err)
1083 goto out_unlock;
1084
1085 /*
1086 * We've written the data and are going to read it back to make
1087 * sure it was written correctly.
1088 */
1089 buf1 = kmalloc(aldata_size, GFP_KERNEL);
1090 if (!buf1) {
1091 err = -ENOMEM;
1092 goto out_unlock;
1093 }
1094
1095 cond_resched();
1096
1097 err = ubi_io_read_data(ubi, buf1, to, 0, aldata_size);
1098 if (err) {
1099 if (err != UBI_IO_BITFLIPS)
1100 ubi_warn("cannot read data back from PEB %d",
1101 to);
1102 goto out_unlock;
1103 }
1104
1105 cond_resched();
1106
1107 if (memcmp(buf, buf1, aldata_size)) {
1108 ubi_warn("read data back from PEB %d - it is different",
1109 to);
1110 goto out_unlock;
1111 }
1112 }
1113
1114 ubi_assert(vol->eba_tbl[lnum] == from);
1115 vol->eba_tbl[lnum] = to;
1116
1117 leb_write_unlock(ubi, vol_id, lnum);
1118 kfree(buf);
1119 kfree(buf1);
1120
1121 return 0;
1122
1123out_unlock:
1124 leb_write_unlock(ubi, vol_id, lnum);
1125 kfree(buf);
1126 kfree(buf1);
1127 return err;
1128}
1129
1130/**
1131 * ubi_eba_init_scan - initialize the EBA unit using scanning information.
1132 * @ubi: UBI device description object
1133 * @si: scanning information
1134 *
1135 * This function returns zero in case of success and a negative error code in
1136 * case of failure.
1137 */
1138int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1139{
1140 int i, j, err, num_volumes;
1141 struct ubi_scan_volume *sv;
1142 struct ubi_volume *vol;
1143 struct ubi_scan_leb *seb;
1144 struct rb_node *rb;
1145
1146 dbg_eba("initialize EBA unit");
1147
1148 spin_lock_init(&ubi->ltree_lock);
1149 ubi->ltree = RB_ROOT;
1150
1151 if (ubi_devices_cnt == 0) {
1152 ltree_slab = kmem_cache_create("ubi_ltree_slab",
1153 sizeof(struct ltree_entry), 0,
1154 0, &ltree_entry_ctor, NULL);
1155 if (!ltree_slab)
1156 return -ENOMEM;
1157 }
1158
1159 ubi->global_sqnum = si->max_sqnum + 1;
1160 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1161
1162 for (i = 0; i < num_volumes; i++) {
1163 vol = ubi->volumes[i];
1164 if (!vol)
1165 continue;
1166
1167 cond_resched();
1168
1169 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1170 GFP_KERNEL);
1171 if (!vol->eba_tbl) {
1172 err = -ENOMEM;
1173 goto out_free;
1174 }
1175
1176 for (j = 0; j < vol->reserved_pebs; j++)
1177 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1178
1179 sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
1180 if (!sv)
1181 continue;
1182
1183 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
1184 if (seb->lnum >= vol->reserved_pebs)
1185 /*
1186 * This may happen in case of an unclean reboot
1187 * during re-size.
1188 */
1189 ubi_scan_move_to_list(sv, seb, &si->erase);
1190 vol->eba_tbl[seb->lnum] = seb->pnum;
1191 }
1192 }
1193
1194 if (ubi->bad_allowed) {
1195 ubi_calculate_reserved(ubi);
1196
1197 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1198 /* No enough free physical eraseblocks */
1199 ubi->beb_rsvd_pebs = ubi->avail_pebs;
1200 ubi_warn("cannot reserve enough PEBs for bad PEB "
1201 "handling, reserved %d, need %d",
1202 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
1203 } else
1204 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1205
1206 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1207 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1208 }
1209
1210 dbg_eba("EBA unit is initialized");
1211 return 0;
1212
1213out_free:
1214 for (i = 0; i < num_volumes; i++) {
1215 if (!ubi->volumes[i])
1216 continue;
1217 kfree(ubi->volumes[i]->eba_tbl);
1218 }
1219 if (ubi_devices_cnt == 0)
1220 kmem_cache_destroy(ltree_slab);
1221 return err;
1222}
1223
1224/**
1225 * ubi_eba_close - close EBA unit.
1226 * @ubi: UBI device description object
1227 */
1228void ubi_eba_close(const struct ubi_device *ubi)
1229{
1230 int i, num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1231
1232 dbg_eba("close EBA unit");
1233
1234 for (i = 0; i < num_volumes; i++) {
1235 if (!ubi->volumes[i])
1236 continue;
1237 kfree(ubi->volumes[i]->eba_tbl);
1238 }
1239 if (ubi_devices_cnt == 1)
1240 kmem_cache_destroy(ltree_slab);
1241}
diff --git a/drivers/mtd/ubi/gluebi.c b/drivers/mtd/ubi/gluebi.c
new file mode 100644
index 000000000000..c8bbfd1e67ab
--- /dev/null
+++ b/drivers/mtd/ubi/gluebi.c
@@ -0,0 +1,324 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём), Joern Engel
19 */
20
21/*
22 * This file includes implementation of fake MTD devices for each UBI volume.
23 * This sounds strange, but it is in fact quite useful to make MTD-oriented
24 * software (including all the legacy software) to work on top of UBI.
25 *
26 * Gluebi emulates MTD devices of "MTD_UBIVOLUME" type. Their minimal I/O unit
27 * size (mtd->writesize) is equivalent to the UBI minimal I/O unit. The
28 * eraseblock size is equivalent to the logical eraseblock size of the volume.
29 */
30
31#include <asm/div64.h>
32#include "ubi.h"
33
34/**
35 * gluebi_get_device - get MTD device reference.
36 * @mtd: the MTD device description object
37 *
38 * This function is called every time the MTD device is being opened and
39 * implements the MTD get_device() operation. Returns zero in case of success
40 * and a negative error code in case of failure.
41 */
42static int gluebi_get_device(struct mtd_info *mtd)
43{
44 struct ubi_volume *vol;
45
46 vol = container_of(mtd, struct ubi_volume, gluebi_mtd);
47
48 /*
49 * We do not introduce locks for gluebi reference count because the
50 * get_device()/put_device() calls are already serialized at MTD.
51 */
52 if (vol->gluebi_refcount > 0) {
53 /*
54 * The MTD device is already referenced and this is just one
55 * more reference. MTD allows many users to open the same
56 * volume simultaneously and do not distinguish between
57 * readers/writers/exclusive openers as UBI does. So we do not
58 * open the UBI volume again - just increase the reference
59 * counter and return.
60 */
61 vol->gluebi_refcount += 1;
62 return 0;
63 }
64
65 /*
66 * This is the first reference to this UBI volume via the MTD device
67 * interface. Open the corresponding volume in read-write mode.
68 */
69 vol->gluebi_desc = ubi_open_volume(vol->ubi->ubi_num, vol->vol_id,
70 UBI_READWRITE);
71 if (IS_ERR(vol->gluebi_desc))
72 return PTR_ERR(vol->gluebi_desc);
73 vol->gluebi_refcount += 1;
74 return 0;
75}
76
77/**
78 * gluebi_put_device - put MTD device reference.
79 * @mtd: the MTD device description object
80 *
81 * This function is called every time the MTD device is being put. Returns
82 * zero in case of success and a negative error code in case of failure.
83 */
84static void gluebi_put_device(struct mtd_info *mtd)
85{
86 struct ubi_volume *vol;
87
88 vol = container_of(mtd, struct ubi_volume, gluebi_mtd);
89 vol->gluebi_refcount -= 1;
90 ubi_assert(vol->gluebi_refcount >= 0);
91 if (vol->gluebi_refcount == 0)
92 ubi_close_volume(vol->gluebi_desc);
93}
94
95/**
96 * gluebi_read - read operation of emulated MTD devices.
97 * @mtd: MTD device description object
98 * @from: absolute offset from where to read
99 * @len: how many bytes to read
100 * @retlen: count of read bytes is returned here
101 * @buf: buffer to store the read data
102 *
103 * This function returns zero in case of success and a negative error code in
104 * case of failure.
105 */
106static int gluebi_read(struct mtd_info *mtd, loff_t from, size_t len,
107 size_t *retlen, unsigned char *buf)
108{
109 int err = 0, lnum, offs, total_read;
110 struct ubi_volume *vol;
111 struct ubi_device *ubi;
112 uint64_t tmp = from;
113
114 dbg_msg("read %zd bytes from offset %lld", len, from);
115
116 if (len < 0 || from < 0 || from + len > mtd->size)
117 return -EINVAL;
118
119 vol = container_of(mtd, struct ubi_volume, gluebi_mtd);
120 ubi = vol->ubi;
121
122 offs = do_div(tmp, mtd->erasesize);
123 lnum = tmp;
124
125 total_read = len;
126 while (total_read) {
127 size_t to_read = mtd->erasesize - offs;
128
129 if (to_read > total_read)
130 to_read = total_read;
131
132 err = ubi_eba_read_leb(ubi, vol->vol_id, lnum, buf, offs,
133 to_read, 0);
134 if (err)
135 break;
136
137 lnum += 1;
138 offs = 0;
139 total_read -= to_read;
140 buf += to_read;
141 }
142
143 *retlen = len - total_read;
144 return err;
145}
146
147/**
148 * gluebi_write - write operation of emulated MTD devices.
149 * @mtd: MTD device description object
150 * @to: absolute offset where to write
151 * @len: how many bytes to write
152 * @retlen: count of written bytes is returned here
153 * @buf: buffer with data to write
154 *
155 * This function returns zero in case of success and a negative error code in
156 * case of failure.
157 */
158static int gluebi_write(struct mtd_info *mtd, loff_t to, size_t len,
159 size_t *retlen, const u_char *buf)
160{
161 int err = 0, lnum, offs, total_written;
162 struct ubi_volume *vol;
163 struct ubi_device *ubi;
164 uint64_t tmp = to;
165
166 dbg_msg("write %zd bytes to offset %lld", len, to);
167
168 if (len < 0 || to < 0 || len + to > mtd->size)
169 return -EINVAL;
170
171 vol = container_of(mtd, struct ubi_volume, gluebi_mtd);
172 ubi = vol->ubi;
173
174 if (ubi->ro_mode)
175 return -EROFS;
176
177 offs = do_div(tmp, mtd->erasesize);
178 lnum = tmp;
179
180 if (len % mtd->writesize || offs % mtd->writesize)
181 return -EINVAL;
182
183 total_written = len;
184 while (total_written) {
185 size_t to_write = mtd->erasesize - offs;
186
187 if (to_write > total_written)
188 to_write = total_written;
189
190 err = ubi_eba_write_leb(ubi, vol->vol_id, lnum, buf, offs,
191 to_write, UBI_UNKNOWN);
192 if (err)
193 break;
194
195 lnum += 1;
196 offs = 0;
197 total_written -= to_write;
198 buf += to_write;
199 }
200
201 *retlen = len - total_written;
202 return err;
203}
204
205/**
206 * gluebi_erase - erase operation of emulated MTD devices.
207 * @mtd: the MTD device description object
208 * @instr: the erase operation description
209 *
210 * This function calls the erase callback when finishes. Returns zero in case
211 * of success and a negative error code in case of failure.
212 */
213static int gluebi_erase(struct mtd_info *mtd, struct erase_info *instr)
214{
215 int err, i, lnum, count;
216 struct ubi_volume *vol;
217 struct ubi_device *ubi;
218
219 dbg_msg("erase %u bytes at offset %u", instr->len, instr->addr);
220
221 if (instr->addr < 0 || instr->addr > mtd->size - mtd->erasesize)
222 return -EINVAL;
223
224 if (instr->len < 0 || instr->addr + instr->len > mtd->size)
225 return -EINVAL;
226
227 if (instr->addr % mtd->writesize || instr->len % mtd->writesize)
228 return -EINVAL;
229
230 lnum = instr->addr / mtd->erasesize;
231 count = instr->len / mtd->erasesize;
232
233 vol = container_of(mtd, struct ubi_volume, gluebi_mtd);
234 ubi = vol->ubi;
235
236 if (ubi->ro_mode)
237 return -EROFS;
238
239 for (i = 0; i < count; i++) {
240 err = ubi_eba_unmap_leb(ubi, vol->vol_id, lnum + i);
241 if (err)
242 goto out_err;
243 }
244
245 /*
246 * MTD erase operations are synchronous, so we have to make sure the
247 * physical eraseblock is wiped out.
248 */
249 err = ubi_wl_flush(ubi);
250 if (err)
251 goto out_err;
252
253 instr->state = MTD_ERASE_DONE;
254 mtd_erase_callback(instr);
255 return 0;
256
257out_err:
258 instr->state = MTD_ERASE_FAILED;
259 instr->fail_addr = lnum * mtd->erasesize;
260 return err;
261}
262
263/**
264 * ubi_create_gluebi - initialize gluebi for an UBI volume.
265 * @ubi: UBI device description object
266 * @vol: volume description object
267 *
268 * This function is called when an UBI volume is created in order to create
269 * corresponding fake MTD device. Returns zero in case of success and a
270 * negative error code in case of failure.
271 */
272int ubi_create_gluebi(struct ubi_device *ubi, struct ubi_volume *vol)
273{
274 int err;
275 struct mtd_info *mtd = &vol->gluebi_mtd;
276
277 mtd->name = kmemdup(vol->name, vol->name_len + 1, GFP_KERNEL);
278 if (!mtd->name)
279 return -ENOMEM;
280
281 mtd->type = MTD_UBIVOLUME;
282 if (!ubi->ro_mode)
283 mtd->flags = MTD_WRITEABLE;
284 mtd->writesize = ubi->min_io_size;
285 mtd->owner = THIS_MODULE;
286 mtd->size = vol->usable_leb_size * vol->reserved_pebs;
287 mtd->erasesize = vol->usable_leb_size;
288 mtd->read = gluebi_read;
289 mtd->write = gluebi_write;
290 mtd->erase = gluebi_erase;
291 mtd->get_device = gluebi_get_device;
292 mtd->put_device = gluebi_put_device;
293
294 if (add_mtd_device(mtd)) {
295 ubi_err("cannot not add MTD device\n");
296 kfree(mtd->name);
297 return -ENFILE;
298 }
299
300 dbg_msg("added mtd%d (\"%s\"), size %u, EB size %u",
301 mtd->index, mtd->name, mtd->size, mtd->erasesize);
302 return 0;
303}
304
305/**
306 * ubi_destroy_gluebi - close gluebi for an UBI volume.
307 * @vol: volume description object
308 *
309 * This function is called when an UBI volume is removed in order to remove
310 * corresponding fake MTD device. Returns zero in case of success and a
311 * negative error code in case of failure.
312 */
313int ubi_destroy_gluebi(struct ubi_volume *vol)
314{
315 int err;
316 struct mtd_info *mtd = &vol->gluebi_mtd;
317
318 dbg_msg("remove mtd%d", mtd->index);
319 err = del_mtd_device(mtd);
320 if (err)
321 return err;
322 kfree(mtd->name);
323 return 0;
324}
diff --git a/drivers/mtd/ubi/io.c b/drivers/mtd/ubi/io.c
new file mode 100644
index 000000000000..438914d05151
--- /dev/null
+++ b/drivers/mtd/ubi/io.c
@@ -0,0 +1,1259 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
3 * Copyright (c) Nokia Corporation, 2006, 2007
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 * the GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 *
19 * Author: Artem Bityutskiy (Битюцкий Артём)
20 */
21
22/*
23 * UBI input/output unit.
24 *
25 * This unit provides a uniform way to work with all kinds of the underlying
26 * MTD devices. It also implements handy functions for reading and writing UBI
27 * headers.
28 *
29 * We are trying to have a paranoid mindset and not to trust to what we read
30 * from the flash media in order to be more secure and robust. So this unit
31 * validates every single header it reads from the flash media.
32 *
33 * Some words about how the eraseblock headers are stored.
34 *
35 * The erase counter header is always stored at offset zero. By default, the
36 * VID header is stored after the EC header at the closest aligned offset
37 * (i.e. aligned to the minimum I/O unit size). Data starts next to the VID
38 * header at the closest aligned offset. But this default layout may be
39 * changed. For example, for different reasons (e.g., optimization) UBI may be
40 * asked to put the VID header at further offset, and even at an unaligned
41 * offset. Of course, if the offset of the VID header is unaligned, UBI adds
42 * proper padding in front of it. Data offset may also be changed but it has to
43 * be aligned.
44 *
45 * About minimal I/O units. In general, UBI assumes flash device model where
46 * there is only one minimal I/O unit size. E.g., in case of NOR flash it is 1,
47 * in case of NAND flash it is a NAND page, etc. This is reported by MTD in the
48 * @ubi->mtd->writesize field. But as an exception, UBI admits of using another
49 * (smaller) minimal I/O unit size for EC and VID headers to make it possible
50 * to do different optimizations.
51 *
52 * This is extremely useful in case of NAND flashes which admit of several
53 * write operations to one NAND page. In this case UBI can fit EC and VID
54 * headers at one NAND page. Thus, UBI may use "sub-page" size as the minimal
55 * I/O unit for the headers (the @ubi->hdrs_min_io_size field). But it still
56 * reports NAND page size (@ubi->min_io_size) as a minimal I/O unit for the UBI
57 * users.
58 *
59 * Example: some Samsung NANDs with 2KiB pages allow 4x 512-byte writes, so
60 * although the minimal I/O unit is 2K, UBI uses 512 bytes for EC and VID
61 * headers.
62 *
63 * Q: why not just to treat sub-page as a minimal I/O unit of this flash
64 * device, e.g., make @ubi->min_io_size = 512 in the example above?
65 *
66 * A: because when writing a sub-page, MTD still writes a full 2K page but the
67 * bytes which are no relevant to the sub-page are 0xFF. So, basically, writing
68 * 4x512 sub-pages is 4 times slower then writing one 2KiB NAND page. Thus, we
69 * prefer to use sub-pages only for EV and VID headers.
70 *
71 * As it was noted above, the VID header may start at a non-aligned offset.
72 * For example, in case of a 2KiB page NAND flash with a 512 bytes sub-page,
73 * the VID header may reside at offset 1984 which is the last 64 bytes of the
74 * last sub-page (EC header is always at offset zero). This causes some
75 * difficulties when reading and writing VID headers.
76 *
77 * Suppose we have a 64-byte buffer and we read a VID header at it. We change
78 * the data and want to write this VID header out. As we can only write in
79 * 512-byte chunks, we have to allocate one more buffer and copy our VID header
80 * to offset 448 of this buffer.
81 *
82 * The I/O unit does the following trick in order to avoid this extra copy.
83 * It always allocates a @ubi->vid_hdr_alsize bytes buffer for the VID header
84 * and returns a pointer to offset @ubi->vid_hdr_shift of this buffer. When the
85 * VID header is being written out, it shifts the VID header pointer back and
86 * writes the whole sub-page.
87 */
88
89#include <linux/crc32.h>
90#include <linux/err.h>
91#include "ubi.h"
92
93#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
94static int paranoid_check_not_bad(const struct ubi_device *ubi, int pnum);
95static int paranoid_check_peb_ec_hdr(const struct ubi_device *ubi, int pnum);
96static int paranoid_check_ec_hdr(const struct ubi_device *ubi, int pnum,
97 const struct ubi_ec_hdr *ec_hdr);
98static int paranoid_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum);
99static int paranoid_check_vid_hdr(const struct ubi_device *ubi, int pnum,
100 const struct ubi_vid_hdr *vid_hdr);
101static int paranoid_check_all_ff(const struct ubi_device *ubi, int pnum,
102 int offset, int len);
103#else
104#define paranoid_check_not_bad(ubi, pnum) 0
105#define paranoid_check_peb_ec_hdr(ubi, pnum) 0
106#define paranoid_check_ec_hdr(ubi, pnum, ec_hdr) 0
107#define paranoid_check_peb_vid_hdr(ubi, pnum) 0
108#define paranoid_check_vid_hdr(ubi, pnum, vid_hdr) 0
109#define paranoid_check_all_ff(ubi, pnum, offset, len) 0
110#endif
111
112/**
113 * ubi_io_read - read data from a physical eraseblock.
114 * @ubi: UBI device description object
115 * @buf: buffer where to store the read data
116 * @pnum: physical eraseblock number to read from
117 * @offset: offset within the physical eraseblock from where to read
118 * @len: how many bytes to read
119 *
120 * This function reads data from offset @offset of physical eraseblock @pnum
121 * and stores the read data in the @buf buffer. The following return codes are
122 * possible:
123 *
124 * o %0 if all the requested data were successfully read;
125 * o %UBI_IO_BITFLIPS if all the requested data were successfully read, but
126 * correctable bit-flips were detected; this is harmless but may indicate
127 * that this eraseblock may become bad soon (but do not have to);
128 * o %-EBADMSG if the MTD subsystem reported about data data integrity
129 * problems, for example it can me an ECC error in case of NAND; this most
130 * probably means that the data is corrupted;
131 * o %-EIO if some I/O error occurred;
132 * o other negative error codes in case of other errors.
133 */
134int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
135 int len)
136{
137 int err, retries = 0;
138 size_t read;
139 loff_t addr;
140
141 dbg_io("read %d bytes from PEB %d:%d", len, pnum, offset);
142
143 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
144 ubi_assert(offset >= 0 && offset + len <= ubi->peb_size);
145 ubi_assert(len > 0);
146
147 err = paranoid_check_not_bad(ubi, pnum);
148 if (err)
149 return err > 0 ? -EINVAL : err;
150
151 addr = (loff_t)pnum * ubi->peb_size + offset;
152retry:
153 err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf);
154 if (err) {
155 if (err == -EUCLEAN) {
156 /*
157 * -EUCLEAN is reported if there was a bit-flip which
158 * was corrected, so this is harmless.
159 */
160 ubi_msg("fixable bit-flip detected at PEB %d", pnum);
161 ubi_assert(len == read);
162 return UBI_IO_BITFLIPS;
163 }
164
165 if (read != len && retries++ < UBI_IO_RETRIES) {
166 dbg_io("error %d while reading %d bytes from PEB %d:%d, "
167 "read only %zd bytes, retry",
168 err, len, pnum, offset, read);
169 yield();
170 goto retry;
171 }
172
173 ubi_err("error %d while reading %d bytes from PEB %d:%d, "
174 "read %zd bytes", err, len, pnum, offset, read);
175 ubi_dbg_dump_stack();
176 } else {
177 ubi_assert(len == read);
178
179 if (ubi_dbg_is_bitflip()) {
180 dbg_msg("bit-flip (emulated)");
181 err = UBI_IO_BITFLIPS;
182 }
183 }
184
185 return err;
186}
187
188/**
189 * ubi_io_write - write data to a physical eraseblock.
190 * @ubi: UBI device description object
191 * @buf: buffer with the data to write
192 * @pnum: physical eraseblock number to write to
193 * @offset: offset within the physical eraseblock where to write
194 * @len: how many bytes to write
195 *
196 * This function writes @len bytes of data from buffer @buf to offset @offset
197 * of physical eraseblock @pnum. If all the data were successfully written,
198 * zero is returned. If an error occurred, this function returns a negative
199 * error code. If %-EIO is returned, the physical eraseblock most probably went
200 * bad.
201 *
202 * Note, in case of an error, it is possible that something was still written
203 * to the flash media, but may be some garbage.
204 */
205int ubi_io_write(const struct ubi_device *ubi, const void *buf, int pnum,
206 int offset, int len)
207{
208 int err;
209 size_t written;
210 loff_t addr;
211
212 dbg_io("write %d bytes to PEB %d:%d", len, pnum, offset);
213
214 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
215 ubi_assert(offset >= 0 && offset + len <= ubi->peb_size);
216 ubi_assert(offset % ubi->hdrs_min_io_size == 0);
217 ubi_assert(len > 0 && len % ubi->hdrs_min_io_size == 0);
218
219 if (ubi->ro_mode) {
220 ubi_err("read-only mode");
221 return -EROFS;
222 }
223
224 /* The below has to be compiled out if paranoid checks are disabled */
225
226 err = paranoid_check_not_bad(ubi, pnum);
227 if (err)
228 return err > 0 ? -EINVAL : err;
229
230 /* The area we are writing to has to contain all 0xFF bytes */
231 err = paranoid_check_all_ff(ubi, pnum, offset, len);
232 if (err)
233 return err > 0 ? -EINVAL : err;
234
235 if (offset >= ubi->leb_start) {
236 /*
237 * We write to the data area of the physical eraseblock. Make
238 * sure it has valid EC and VID headers.
239 */
240 err = paranoid_check_peb_ec_hdr(ubi, pnum);
241 if (err)
242 return err > 0 ? -EINVAL : err;
243 err = paranoid_check_peb_vid_hdr(ubi, pnum);
244 if (err)
245 return err > 0 ? -EINVAL : err;
246 }
247
248 if (ubi_dbg_is_write_failure()) {
249 dbg_err("cannot write %d bytes to PEB %d:%d "
250 "(emulated)", len, pnum, offset);
251 ubi_dbg_dump_stack();
252 return -EIO;
253 }
254
255 addr = (loff_t)pnum * ubi->peb_size + offset;
256 err = ubi->mtd->write(ubi->mtd, addr, len, &written, buf);
257 if (err) {
258 ubi_err("error %d while writing %d bytes to PEB %d:%d, written"
259 " %zd bytes", err, len, pnum, offset, written);
260 ubi_dbg_dump_stack();
261 } else
262 ubi_assert(written == len);
263
264 return err;
265}
266
267/**
268 * erase_callback - MTD erasure call-back.
269 * @ei: MTD erase information object.
270 *
271 * Note, even though MTD erase interface is asynchronous, all the current
272 * implementations are synchronous anyway.
273 */
274static void erase_callback(struct erase_info *ei)
275{
276 wake_up_interruptible((wait_queue_head_t *)ei->priv);
277}
278
279/**
280 * do_sync_erase - synchronously erase a physical eraseblock.
281 * @ubi: UBI device description object
282 * @pnum: the physical eraseblock number to erase
283 *
284 * This function synchronously erases physical eraseblock @pnum and returns
285 * zero in case of success and a negative error code in case of failure. If
286 * %-EIO is returned, the physical eraseblock most probably went bad.
287 */
288static int do_sync_erase(const struct ubi_device *ubi, int pnum)
289{
290 int err, retries = 0;
291 struct erase_info ei;
292 wait_queue_head_t wq;
293
294 dbg_io("erase PEB %d", pnum);
295
296retry:
297 init_waitqueue_head(&wq);
298 memset(&ei, 0, sizeof(struct erase_info));
299
300 ei.mtd = ubi->mtd;
301 ei.addr = pnum * ubi->peb_size;
302 ei.len = ubi->peb_size;
303 ei.callback = erase_callback;
304 ei.priv = (unsigned long)&wq;
305
306 err = ubi->mtd->erase(ubi->mtd, &ei);
307 if (err) {
308 if (retries++ < UBI_IO_RETRIES) {
309 dbg_io("error %d while erasing PEB %d, retry",
310 err, pnum);
311 yield();
312 goto retry;
313 }
314 ubi_err("cannot erase PEB %d, error %d", pnum, err);
315 ubi_dbg_dump_stack();
316 return err;
317 }
318
319 err = wait_event_interruptible(wq, ei.state == MTD_ERASE_DONE ||
320 ei.state == MTD_ERASE_FAILED);
321 if (err) {
322 ubi_err("interrupted PEB %d erasure", pnum);
323 return -EINTR;
324 }
325
326 if (ei.state == MTD_ERASE_FAILED) {
327 if (retries++ < UBI_IO_RETRIES) {
328 dbg_io("error while erasing PEB %d, retry", pnum);
329 yield();
330 goto retry;
331 }
332 ubi_err("cannot erase PEB %d", pnum);
333 ubi_dbg_dump_stack();
334 return -EIO;
335 }
336
337 err = paranoid_check_all_ff(ubi, pnum, 0, ubi->peb_size);
338 if (err)
339 return err > 0 ? -EINVAL : err;
340
341 if (ubi_dbg_is_erase_failure() && !err) {
342 dbg_err("cannot erase PEB %d (emulated)", pnum);
343 return -EIO;
344 }
345
346 return 0;
347}
348
349/**
350 * check_pattern - check if buffer contains only a certain byte pattern.
351 * @buf: buffer to check
352 * @patt: the pattern to check
353 * @size: buffer size in bytes
354 *
355 * This function returns %1 in there are only @patt bytes in @buf, and %0 if
356 * something else was also found.
357 */
358static int check_pattern(const void *buf, uint8_t patt, int size)
359{
360 int i;
361
362 for (i = 0; i < size; i++)
363 if (((const uint8_t *)buf)[i] != patt)
364 return 0;
365 return 1;
366}
367
368/* Patterns to write to a physical eraseblock when torturing it */
369static uint8_t patterns[] = {0xa5, 0x5a, 0x0};
370
371/**
372 * torture_peb - test a supposedly bad physical eraseblock.
373 * @ubi: UBI device description object
374 * @pnum: the physical eraseblock number to test
375 *
376 * This function returns %-EIO if the physical eraseblock did not pass the
377 * test, a positive number of erase operations done if the test was
378 * successfully passed, and other negative error codes in case of other errors.
379 */
380static int torture_peb(const struct ubi_device *ubi, int pnum)
381{
382 void *buf;
383 int err, i, patt_count;
384
385 buf = kmalloc(ubi->peb_size, GFP_KERNEL);
386 if (!buf)
387 return -ENOMEM;
388
389 patt_count = ARRAY_SIZE(patterns);
390 ubi_assert(patt_count > 0);
391
392 for (i = 0; i < patt_count; i++) {
393 err = do_sync_erase(ubi, pnum);
394 if (err)
395 goto out;
396
397 /* Make sure the PEB contains only 0xFF bytes */
398 err = ubi_io_read(ubi, buf, pnum, 0, ubi->peb_size);
399 if (err)
400 goto out;
401
402 err = check_pattern(buf, 0xFF, ubi->peb_size);
403 if (err == 0) {
404 ubi_err("erased PEB %d, but a non-0xFF byte found",
405 pnum);
406 err = -EIO;
407 goto out;
408 }
409
410 /* Write a pattern and check it */
411 memset(buf, patterns[i], ubi->peb_size);
412 err = ubi_io_write(ubi, buf, pnum, 0, ubi->peb_size);
413 if (err)
414 goto out;
415
416 memset(buf, ~patterns[i], ubi->peb_size);
417 err = ubi_io_read(ubi, buf, pnum, 0, ubi->peb_size);
418 if (err)
419 goto out;
420
421 err = check_pattern(buf, patterns[i], ubi->peb_size);
422 if (err == 0) {
423 ubi_err("pattern %x checking failed for PEB %d",
424 patterns[i], pnum);
425 err = -EIO;
426 goto out;
427 }
428 }
429
430 err = patt_count;
431
432out:
433 if (err == UBI_IO_BITFLIPS || err == -EBADMSG)
434 /*
435 * If a bit-flip or data integrity error was detected, the test
436 * has not passed because it happened on a freshly erased
437 * physical eraseblock which means something is wrong with it.
438 */
439 err = -EIO;
440 kfree(buf);
441 return err;
442}
443
444/**
445 * ubi_io_sync_erase - synchronously erase a physical eraseblock.
446 * @ubi: UBI device description object
447 * @pnum: physical eraseblock number to erase
448 * @torture: if this physical eraseblock has to be tortured
449 *
450 * This function synchronously erases physical eraseblock @pnum. If @torture
451 * flag is not zero, the physical eraseblock is checked by means of writing
452 * different patterns to it and reading them back. If the torturing is enabled,
453 * the physical eraseblock is erased more then once.
454 *
455 * This function returns the number of erasures made in case of success, %-EIO
456 * if the erasure failed or the torturing test failed, and other negative error
457 * codes in case of other errors. Note, %-EIO means that the physical
458 * eraseblock is bad.
459 */
460int ubi_io_sync_erase(const struct ubi_device *ubi, int pnum, int torture)
461{
462 int err, ret = 0;
463
464 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
465
466 err = paranoid_check_not_bad(ubi, pnum);
467 if (err != 0)
468 return err > 0 ? -EINVAL : err;
469
470 if (ubi->ro_mode) {
471 ubi_err("read-only mode");
472 return -EROFS;
473 }
474
475 if (torture) {
476 ret = torture_peb(ubi, pnum);
477 if (ret < 0)
478 return ret;
479 }
480
481 err = do_sync_erase(ubi, pnum);
482 if (err)
483 return err;
484
485 return ret + 1;
486}
487
488/**
489 * ubi_io_is_bad - check if a physical eraseblock is bad.
490 * @ubi: UBI device description object
491 * @pnum: the physical eraseblock number to check
492 *
493 * This function returns a positive number if the physical eraseblock is bad,
494 * zero if not, and a negative error code if an error occurred.
495 */
496int ubi_io_is_bad(const struct ubi_device *ubi, int pnum)
497{
498 struct mtd_info *mtd = ubi->mtd;
499
500 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
501
502 if (ubi->bad_allowed) {
503 int ret;
504
505 ret = mtd->block_isbad(mtd, (loff_t)pnum * ubi->peb_size);
506 if (ret < 0)
507 ubi_err("error %d while checking if PEB %d is bad",
508 ret, pnum);
509 else if (ret)
510 dbg_io("PEB %d is bad", pnum);
511 return ret;
512 }
513
514 return 0;
515}
516
517/**
518 * ubi_io_mark_bad - mark a physical eraseblock as bad.
519 * @ubi: UBI device description object
520 * @pnum: the physical eraseblock number to mark
521 *
522 * This function returns zero in case of success and a negative error code in
523 * case of failure.
524 */
525int ubi_io_mark_bad(const struct ubi_device *ubi, int pnum)
526{
527 int err;
528 struct mtd_info *mtd = ubi->mtd;
529
530 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
531
532 if (ubi->ro_mode) {
533 ubi_err("read-only mode");
534 return -EROFS;
535 }
536
537 if (!ubi->bad_allowed)
538 return 0;
539
540 err = mtd->block_markbad(mtd, (loff_t)pnum * ubi->peb_size);
541 if (err)
542 ubi_err("cannot mark PEB %d bad, error %d", pnum, err);
543 return err;
544}
545
546/**
547 * validate_ec_hdr - validate an erase counter header.
548 * @ubi: UBI device description object
549 * @ec_hdr: the erase counter header to check
550 *
551 * This function returns zero if the erase counter header is OK, and %1 if
552 * not.
553 */
554static int validate_ec_hdr(const struct ubi_device *ubi,
555 const struct ubi_ec_hdr *ec_hdr)
556{
557 long long ec;
558 int vid_hdr_offset, leb_start;
559
560 ec = ubi64_to_cpu(ec_hdr->ec);
561 vid_hdr_offset = ubi32_to_cpu(ec_hdr->vid_hdr_offset);
562 leb_start = ubi32_to_cpu(ec_hdr->data_offset);
563
564 if (ec_hdr->version != UBI_VERSION) {
565 ubi_err("node with incompatible UBI version found: "
566 "this UBI version is %d, image version is %d",
567 UBI_VERSION, (int)ec_hdr->version);
568 goto bad;
569 }
570
571 if (vid_hdr_offset != ubi->vid_hdr_offset) {
572 ubi_err("bad VID header offset %d, expected %d",
573 vid_hdr_offset, ubi->vid_hdr_offset);
574 goto bad;
575 }
576
577 if (leb_start != ubi->leb_start) {
578 ubi_err("bad data offset %d, expected %d",
579 leb_start, ubi->leb_start);
580 goto bad;
581 }
582
583 if (ec < 0 || ec > UBI_MAX_ERASECOUNTER) {
584 ubi_err("bad erase counter %lld", ec);
585 goto bad;
586 }
587
588 return 0;
589
590bad:
591 ubi_err("bad EC header");
592 ubi_dbg_dump_ec_hdr(ec_hdr);
593 ubi_dbg_dump_stack();
594 return 1;
595}
596
597/**
598 * ubi_io_read_ec_hdr - read and check an erase counter header.
599 * @ubi: UBI device description object
600 * @pnum: physical eraseblock to read from
601 * @ec_hdr: a &struct ubi_ec_hdr object where to store the read erase counter
602 * header
603 * @verbose: be verbose if the header is corrupted or was not found
604 *
605 * This function reads erase counter header from physical eraseblock @pnum and
606 * stores it in @ec_hdr. This function also checks CRC checksum of the read
607 * erase counter header. The following codes may be returned:
608 *
609 * o %0 if the CRC checksum is correct and the header was successfully read;
610 * o %UBI_IO_BITFLIPS if the CRC is correct, but bit-flips were detected
611 * and corrected by the flash driver; this is harmless but may indicate that
612 * this eraseblock may become bad soon (but may be not);
613 * o %UBI_IO_BAD_EC_HDR if the erase counter header is corrupted (a CRC error);
614 * o %UBI_IO_PEB_EMPTY if the physical eraseblock is empty;
615 * o a negative error code in case of failure.
616 */
617int ubi_io_read_ec_hdr(const struct ubi_device *ubi, int pnum,
618 struct ubi_ec_hdr *ec_hdr, int verbose)
619{
620 int err, read_err = 0;
621 uint32_t crc, magic, hdr_crc;
622
623 dbg_io("read EC header from PEB %d", pnum);
624 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
625
626 err = ubi_io_read(ubi, ec_hdr, pnum, 0, UBI_EC_HDR_SIZE);
627 if (err) {
628 if (err != UBI_IO_BITFLIPS && err != -EBADMSG)
629 return err;
630
631 /*
632 * We read all the data, but either a correctable bit-flip
633 * occurred, or MTD reported about some data integrity error,
634 * like an ECC error in case of NAND. The former is harmless,
635 * the later may mean that the read data is corrupted. But we
636 * have a CRC check-sum and we will detect this. If the EC
637 * header is still OK, we just report this as there was a
638 * bit-flip.
639 */
640 read_err = err;
641 }
642
643 magic = ubi32_to_cpu(ec_hdr->magic);
644 if (magic != UBI_EC_HDR_MAGIC) {
645 /*
646 * The magic field is wrong. Let's check if we have read all
647 * 0xFF. If yes, this physical eraseblock is assumed to be
648 * empty.
649 *
650 * But if there was a read error, we do not test it for all
651 * 0xFFs. Even if it does contain all 0xFFs, this error
652 * indicates that something is still wrong with this physical
653 * eraseblock and we anyway cannot treat it as empty.
654 */
655 if (read_err != -EBADMSG &&
656 check_pattern(ec_hdr, 0xFF, UBI_EC_HDR_SIZE)) {
657 /* The physical eraseblock is supposedly empty */
658
659 /*
660 * The below is just a paranoid check, it has to be
661 * compiled out if paranoid checks are disabled.
662 */
663 err = paranoid_check_all_ff(ubi, pnum, 0,
664 ubi->peb_size);
665 if (err)
666 return err > 0 ? UBI_IO_BAD_EC_HDR : err;
667
668 if (verbose)
669 ubi_warn("no EC header found at PEB %d, "
670 "only 0xFF bytes", pnum);
671 return UBI_IO_PEB_EMPTY;
672 }
673
674 /*
675 * This is not a valid erase counter header, and these are not
676 * 0xFF bytes. Report that the header is corrupted.
677 */
678 if (verbose) {
679 ubi_warn("bad magic number at PEB %d: %08x instead of "
680 "%08x", pnum, magic, UBI_EC_HDR_MAGIC);
681 ubi_dbg_dump_ec_hdr(ec_hdr);
682 }
683 return UBI_IO_BAD_EC_HDR;
684 }
685
686 crc = crc32(UBI_CRC32_INIT, ec_hdr, UBI_EC_HDR_SIZE_CRC);
687 hdr_crc = ubi32_to_cpu(ec_hdr->hdr_crc);
688
689 if (hdr_crc != crc) {
690 if (verbose) {
691 ubi_warn("bad EC header CRC at PEB %d, calculated %#08x,"
692 " read %#08x", pnum, crc, hdr_crc);
693 ubi_dbg_dump_ec_hdr(ec_hdr);
694 }
695 return UBI_IO_BAD_EC_HDR;
696 }
697
698 /* And of course validate what has just been read from the media */
699 err = validate_ec_hdr(ubi, ec_hdr);
700 if (err) {
701 ubi_err("validation failed for PEB %d", pnum);
702 return -EINVAL;
703 }
704
705 return read_err ? UBI_IO_BITFLIPS : 0;
706}
707
708/**
709 * ubi_io_write_ec_hdr - write an erase counter header.
710 * @ubi: UBI device description object
711 * @pnum: physical eraseblock to write to
712 * @ec_hdr: the erase counter header to write
713 *
714 * This function writes erase counter header described by @ec_hdr to physical
715 * eraseblock @pnum. It also fills most fields of @ec_hdr before writing, so
716 * the caller do not have to fill them. Callers must only fill the @ec_hdr->ec
717 * field.
718 *
719 * This function returns zero in case of success and a negative error code in
720 * case of failure. If %-EIO is returned, the physical eraseblock most probably
721 * went bad.
722 */
723int ubi_io_write_ec_hdr(const struct ubi_device *ubi, int pnum,
724 struct ubi_ec_hdr *ec_hdr)
725{
726 int err;
727 uint32_t crc;
728
729 dbg_io("write EC header to PEB %d", pnum);
730 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
731
732 ec_hdr->magic = cpu_to_ubi32(UBI_EC_HDR_MAGIC);
733 ec_hdr->version = UBI_VERSION;
734 ec_hdr->vid_hdr_offset = cpu_to_ubi32(ubi->vid_hdr_offset);
735 ec_hdr->data_offset = cpu_to_ubi32(ubi->leb_start);
736 crc = crc32(UBI_CRC32_INIT, ec_hdr, UBI_EC_HDR_SIZE_CRC);
737 ec_hdr->hdr_crc = cpu_to_ubi32(crc);
738
739 err = paranoid_check_ec_hdr(ubi, pnum, ec_hdr);
740 if (err)
741 return -EINVAL;
742
743 err = ubi_io_write(ubi, ec_hdr, pnum, 0, ubi->ec_hdr_alsize);
744 return err;
745}
746
747/**
748 * validate_vid_hdr - validate a volume identifier header.
749 * @ubi: UBI device description object
750 * @vid_hdr: the volume identifier header to check
751 *
752 * This function checks that data stored in the volume identifier header
753 * @vid_hdr. Returns zero if the VID header is OK and %1 if not.
754 */
755static int validate_vid_hdr(const struct ubi_device *ubi,
756 const struct ubi_vid_hdr *vid_hdr)
757{
758 int vol_type = vid_hdr->vol_type;
759 int copy_flag = vid_hdr->copy_flag;
760 int vol_id = ubi32_to_cpu(vid_hdr->vol_id);
761 int lnum = ubi32_to_cpu(vid_hdr->lnum);
762 int compat = vid_hdr->compat;
763 int data_size = ubi32_to_cpu(vid_hdr->data_size);
764 int used_ebs = ubi32_to_cpu(vid_hdr->used_ebs);
765 int data_pad = ubi32_to_cpu(vid_hdr->data_pad);
766 int data_crc = ubi32_to_cpu(vid_hdr->data_crc);
767 int usable_leb_size = ubi->leb_size - data_pad;
768
769 if (copy_flag != 0 && copy_flag != 1) {
770 dbg_err("bad copy_flag");
771 goto bad;
772 }
773
774 if (vol_id < 0 || lnum < 0 || data_size < 0 || used_ebs < 0 ||
775 data_pad < 0) {
776 dbg_err("negative values");
777 goto bad;
778 }
779
780 if (vol_id >= UBI_MAX_VOLUMES && vol_id < UBI_INTERNAL_VOL_START) {
781 dbg_err("bad vol_id");
782 goto bad;
783 }
784
785 if (vol_id < UBI_INTERNAL_VOL_START && compat != 0) {
786 dbg_err("bad compat");
787 goto bad;
788 }
789
790 if (vol_id >= UBI_INTERNAL_VOL_START && compat != UBI_COMPAT_DELETE &&
791 compat != UBI_COMPAT_RO && compat != UBI_COMPAT_PRESERVE &&
792 compat != UBI_COMPAT_REJECT) {
793 dbg_err("bad compat");
794 goto bad;
795 }
796
797 if (vol_type != UBI_VID_DYNAMIC && vol_type != UBI_VID_STATIC) {
798 dbg_err("bad vol_type");
799 goto bad;
800 }
801
802 if (data_pad >= ubi->leb_size / 2) {
803 dbg_err("bad data_pad");
804 goto bad;
805 }
806
807 if (vol_type == UBI_VID_STATIC) {
808 /*
809 * Although from high-level point of view static volumes may
810 * contain zero bytes of data, but no VID headers can contain
811 * zero at these fields, because they empty volumes do not have
812 * mapped logical eraseblocks.
813 */
814 if (used_ebs == 0) {
815 dbg_err("zero used_ebs");
816 goto bad;
817 }
818 if (data_size == 0) {
819 dbg_err("zero data_size");
820 goto bad;
821 }
822 if (lnum < used_ebs - 1) {
823 if (data_size != usable_leb_size) {
824 dbg_err("bad data_size");
825 goto bad;
826 }
827 } else if (lnum == used_ebs - 1) {
828 if (data_size == 0) {
829 dbg_err("bad data_size at last LEB");
830 goto bad;
831 }
832 } else {
833 dbg_err("too high lnum");
834 goto bad;
835 }
836 } else {
837 if (copy_flag == 0) {
838 if (data_crc != 0) {
839 dbg_err("non-zero data CRC");
840 goto bad;
841 }
842 if (data_size != 0) {
843 dbg_err("non-zero data_size");
844 goto bad;
845 }
846 } else {
847 if (data_size == 0) {
848 dbg_err("zero data_size of copy");
849 goto bad;
850 }
851 }
852 if (used_ebs != 0) {
853 dbg_err("bad used_ebs");
854 goto bad;
855 }
856 }
857
858 return 0;
859
860bad:
861 ubi_err("bad VID header");
862 ubi_dbg_dump_vid_hdr(vid_hdr);
863 ubi_dbg_dump_stack();
864 return 1;
865}
866
867/**
868 * ubi_io_read_vid_hdr - read and check a volume identifier header.
869 * @ubi: UBI device description object
870 * @pnum: physical eraseblock number to read from
871 * @vid_hdr: &struct ubi_vid_hdr object where to store the read volume
872 * identifier header
873 * @verbose: be verbose if the header is corrupted or wasn't found
874 *
875 * This function reads the volume identifier header from physical eraseblock
876 * @pnum and stores it in @vid_hdr. It also checks CRC checksum of the read
877 * volume identifier header. The following codes may be returned:
878 *
879 * o %0 if the CRC checksum is correct and the header was successfully read;
880 * o %UBI_IO_BITFLIPS if the CRC is correct, but bit-flips were detected
881 * and corrected by the flash driver; this is harmless but may indicate that
882 * this eraseblock may become bad soon;
883 * o %UBI_IO_BAD_VID_HRD if the volume identifier header is corrupted (a CRC
884 * error detected);
885 * o %UBI_IO_PEB_FREE if the physical eraseblock is free (i.e., there is no VID
886 * header there);
887 * o a negative error code in case of failure.
888 */
889int ubi_io_read_vid_hdr(const struct ubi_device *ubi, int pnum,
890 struct ubi_vid_hdr *vid_hdr, int verbose)
891{
892 int err, read_err = 0;
893 uint32_t crc, magic, hdr_crc;
894 void *p;
895
896 dbg_io("read VID header from PEB %d", pnum);
897 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
898
899 p = (char *)vid_hdr - ubi->vid_hdr_shift;
900 err = ubi_io_read(ubi, p, pnum, ubi->vid_hdr_aloffset,
901 ubi->vid_hdr_alsize);
902 if (err) {
903 if (err != UBI_IO_BITFLIPS && err != -EBADMSG)
904 return err;
905
906 /*
907 * We read all the data, but either a correctable bit-flip
908 * occurred, or MTD reported about some data integrity error,
909 * like an ECC error in case of NAND. The former is harmless,
910 * the later may mean the read data is corrupted. But we have a
911 * CRC check-sum and we will identify this. If the VID header is
912 * still OK, we just report this as there was a bit-flip.
913 */
914 read_err = err;
915 }
916
917 magic = ubi32_to_cpu(vid_hdr->magic);
918 if (magic != UBI_VID_HDR_MAGIC) {
919 /*
920 * If we have read all 0xFF bytes, the VID header probably does
921 * not exist and the physical eraseblock is assumed to be free.
922 *
923 * But if there was a read error, we do not test the data for
924 * 0xFFs. Even if it does contain all 0xFFs, this error
925 * indicates that something is still wrong with this physical
926 * eraseblock and it cannot be regarded as free.
927 */
928 if (read_err != -EBADMSG &&
929 check_pattern(vid_hdr, 0xFF, UBI_VID_HDR_SIZE)) {
930 /* The physical eraseblock is supposedly free */
931
932 /*
933 * The below is just a paranoid check, it has to be
934 * compiled out if paranoid checks are disabled.
935 */
936 err = paranoid_check_all_ff(ubi, pnum, ubi->leb_start,
937 ubi->leb_size);
938 if (err)
939 return err > 0 ? UBI_IO_BAD_VID_HDR : err;
940
941 if (verbose)
942 ubi_warn("no VID header found at PEB %d, "
943 "only 0xFF bytes", pnum);
944 return UBI_IO_PEB_FREE;
945 }
946
947 /*
948 * This is not a valid VID header, and these are not 0xFF
949 * bytes. Report that the header is corrupted.
950 */
951 if (verbose) {
952 ubi_warn("bad magic number at PEB %d: %08x instead of "
953 "%08x", pnum, magic, UBI_VID_HDR_MAGIC);
954 ubi_dbg_dump_vid_hdr(vid_hdr);
955 }
956 return UBI_IO_BAD_VID_HDR;
957 }
958
959 crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_VID_HDR_SIZE_CRC);
960 hdr_crc = ubi32_to_cpu(vid_hdr->hdr_crc);
961
962 if (hdr_crc != crc) {
963 if (verbose) {
964 ubi_warn("bad CRC at PEB %d, calculated %#08x, "
965 "read %#08x", pnum, crc, hdr_crc);
966 ubi_dbg_dump_vid_hdr(vid_hdr);
967 }
968 return UBI_IO_BAD_VID_HDR;
969 }
970
971 /* Validate the VID header that we have just read */
972 err = validate_vid_hdr(ubi, vid_hdr);
973 if (err) {
974 ubi_err("validation failed for PEB %d", pnum);
975 return -EINVAL;
976 }
977
978 return read_err ? UBI_IO_BITFLIPS : 0;
979}
980
981/**
982 * ubi_io_write_vid_hdr - write a volume identifier header.
983 * @ubi: UBI device description object
984 * @pnum: the physical eraseblock number to write to
985 * @vid_hdr: the volume identifier header to write
986 *
987 * This function writes the volume identifier header described by @vid_hdr to
988 * physical eraseblock @pnum. This function automatically fills the
989 * @vid_hdr->magic and the @vid_hdr->version fields, as well as calculates
990 * header CRC checksum and stores it at vid_hdr->hdr_crc.
991 *
992 * This function returns zero in case of success and a negative error code in
993 * case of failure. If %-EIO is returned, the physical eraseblock probably went
994 * bad.
995 */
996int ubi_io_write_vid_hdr(const struct ubi_device *ubi, int pnum,
997 struct ubi_vid_hdr *vid_hdr)
998{
999 int err;
1000 uint32_t crc;
1001 void *p;
1002
1003 dbg_io("write VID header to PEB %d", pnum);
1004 ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
1005
1006 err = paranoid_check_peb_ec_hdr(ubi, pnum);
1007 if (err)
1008 return err > 0 ? -EINVAL: err;
1009
1010 vid_hdr->magic = cpu_to_ubi32(UBI_VID_HDR_MAGIC);
1011 vid_hdr->version = UBI_VERSION;
1012 crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_VID_HDR_SIZE_CRC);
1013 vid_hdr->hdr_crc = cpu_to_ubi32(crc);
1014
1015 err = paranoid_check_vid_hdr(ubi, pnum, vid_hdr);
1016 if (err)
1017 return -EINVAL;
1018
1019 p = (char *)vid_hdr - ubi->vid_hdr_shift;
1020 err = ubi_io_write(ubi, p, pnum, ubi->vid_hdr_aloffset,
1021 ubi->vid_hdr_alsize);
1022 return err;
1023}
1024
1025#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
1026
1027/**
1028 * paranoid_check_not_bad - ensure that a physical eraseblock is not bad.
1029 * @ubi: UBI device description object
1030 * @pnum: physical eraseblock number to check
1031 *
1032 * This function returns zero if the physical eraseblock is good, a positive
1033 * number if it is bad and a negative error code if an error occurred.
1034 */
1035static int paranoid_check_not_bad(const struct ubi_device *ubi, int pnum)
1036{
1037 int err;
1038
1039 err = ubi_io_is_bad(ubi, pnum);
1040 if (!err)
1041 return err;
1042
1043 ubi_err("paranoid check failed for PEB %d", pnum);
1044 ubi_dbg_dump_stack();
1045 return err;
1046}
1047
1048/**
1049 * paranoid_check_ec_hdr - check if an erase counter header is all right.
1050 * @ubi: UBI device description object
1051 * @pnum: physical eraseblock number the erase counter header belongs to
1052 * @ec_hdr: the erase counter header to check
1053 *
1054 * This function returns zero if the erase counter header contains valid
1055 * values, and %1 if not.
1056 */
1057static int paranoid_check_ec_hdr(const struct ubi_device *ubi, int pnum,
1058 const struct ubi_ec_hdr *ec_hdr)
1059{
1060 int err;
1061 uint32_t magic;
1062
1063 magic = ubi32_to_cpu(ec_hdr->magic);
1064 if (magic != UBI_EC_HDR_MAGIC) {
1065 ubi_err("bad magic %#08x, must be %#08x",
1066 magic, UBI_EC_HDR_MAGIC);
1067 goto fail;
1068 }
1069
1070 err = validate_ec_hdr(ubi, ec_hdr);
1071 if (err) {
1072 ubi_err("paranoid check failed for PEB %d", pnum);
1073 goto fail;
1074 }
1075
1076 return 0;
1077
1078fail:
1079 ubi_dbg_dump_ec_hdr(ec_hdr);
1080 ubi_dbg_dump_stack();
1081 return 1;
1082}
1083
1084/**
1085 * paranoid_check_peb_ec_hdr - check that the erase counter header of a
1086 * physical eraseblock is in-place and is all right.
1087 * @ubi: UBI device description object
1088 * @pnum: the physical eraseblock number to check
1089 *
1090 * This function returns zero if the erase counter header is all right, %1 if
1091 * not, and a negative error code if an error occurred.
1092 */
1093static int paranoid_check_peb_ec_hdr(const struct ubi_device *ubi, int pnum)
1094{
1095 int err;
1096 uint32_t crc, hdr_crc;
1097 struct ubi_ec_hdr *ec_hdr;
1098
1099 ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
1100 if (!ec_hdr)
1101 return -ENOMEM;
1102
1103 err = ubi_io_read(ubi, ec_hdr, pnum, 0, UBI_EC_HDR_SIZE);
1104 if (err && err != UBI_IO_BITFLIPS && err != -EBADMSG)
1105 goto exit;
1106
1107 crc = crc32(UBI_CRC32_INIT, ec_hdr, UBI_EC_HDR_SIZE_CRC);
1108 hdr_crc = ubi32_to_cpu(ec_hdr->hdr_crc);
1109 if (hdr_crc != crc) {
1110 ubi_err("bad CRC, calculated %#08x, read %#08x", crc, hdr_crc);
1111 ubi_err("paranoid check failed for PEB %d", pnum);
1112 ubi_dbg_dump_ec_hdr(ec_hdr);
1113 ubi_dbg_dump_stack();
1114 err = 1;
1115 goto exit;
1116 }
1117
1118 err = paranoid_check_ec_hdr(ubi, pnum, ec_hdr);
1119
1120exit:
1121 kfree(ec_hdr);
1122 return err;
1123}
1124
1125/**
1126 * paranoid_check_vid_hdr - check that a volume identifier header is all right.
1127 * @ubi: UBI device description object
1128 * @pnum: physical eraseblock number the volume identifier header belongs to
1129 * @vid_hdr: the volume identifier header to check
1130 *
1131 * This function returns zero if the volume identifier header is all right, and
1132 * %1 if not.
1133 */
1134static int paranoid_check_vid_hdr(const struct ubi_device *ubi, int pnum,
1135 const struct ubi_vid_hdr *vid_hdr)
1136{
1137 int err;
1138 uint32_t magic;
1139
1140 magic = ubi32_to_cpu(vid_hdr->magic);
1141 if (magic != UBI_VID_HDR_MAGIC) {
1142 ubi_err("bad VID header magic %#08x at PEB %d, must be %#08x",
1143 magic, pnum, UBI_VID_HDR_MAGIC);
1144 goto fail;
1145 }
1146
1147 err = validate_vid_hdr(ubi, vid_hdr);
1148 if (err) {
1149 ubi_err("paranoid check failed for PEB %d", pnum);
1150 goto fail;
1151 }
1152
1153 return err;
1154
1155fail:
1156 ubi_err("paranoid check failed for PEB %d", pnum);
1157 ubi_dbg_dump_vid_hdr(vid_hdr);
1158 ubi_dbg_dump_stack();
1159 return 1;
1160
1161}
1162
1163/**
1164 * paranoid_check_peb_vid_hdr - check that the volume identifier header of a
1165 * physical eraseblock is in-place and is all right.
1166 * @ubi: UBI device description object
1167 * @pnum: the physical eraseblock number to check
1168 *
1169 * This function returns zero if the volume identifier header is all right,
1170 * %1 if not, and a negative error code if an error occurred.
1171 */
1172static int paranoid_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum)
1173{
1174 int err;
1175 uint32_t crc, hdr_crc;
1176 struct ubi_vid_hdr *vid_hdr;
1177 void *p;
1178
1179 vid_hdr = ubi_zalloc_vid_hdr(ubi);
1180 if (!vid_hdr)
1181 return -ENOMEM;
1182
1183 p = (char *)vid_hdr - ubi->vid_hdr_shift;
1184 err = ubi_io_read(ubi, p, pnum, ubi->vid_hdr_aloffset,
1185 ubi->vid_hdr_alsize);
1186 if (err && err != UBI_IO_BITFLIPS && err != -EBADMSG)
1187 goto exit;
1188
1189 crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_EC_HDR_SIZE_CRC);
1190 hdr_crc = ubi32_to_cpu(vid_hdr->hdr_crc);
1191 if (hdr_crc != crc) {
1192 ubi_err("bad VID header CRC at PEB %d, calculated %#08x, "
1193 "read %#08x", pnum, crc, hdr_crc);
1194 ubi_err("paranoid check failed for PEB %d", pnum);
1195 ubi_dbg_dump_vid_hdr(vid_hdr);
1196 ubi_dbg_dump_stack();
1197 err = 1;
1198 goto exit;
1199 }
1200
1201 err = paranoid_check_vid_hdr(ubi, pnum, vid_hdr);
1202
1203exit:
1204 ubi_free_vid_hdr(ubi, vid_hdr);
1205 return err;
1206}
1207
1208/**
1209 * paranoid_check_all_ff - check that a region of flash is empty.
1210 * @ubi: UBI device description object
1211 * @pnum: the physical eraseblock number to check
1212 * @offset: the starting offset within the physical eraseblock to check
1213 * @len: the length of the region to check
1214 *
1215 * This function returns zero if only 0xFF bytes are present at offset
1216 * @offset of the physical eraseblock @pnum, %1 if not, and a negative error
1217 * code if an error occurred.
1218 */
1219static int paranoid_check_all_ff(const struct ubi_device *ubi, int pnum,
1220 int offset, int len)
1221{
1222 size_t read;
1223 int err;
1224 void *buf;
1225 loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
1226
1227 buf = kzalloc(len, GFP_KERNEL);
1228 if (!buf)
1229 return -ENOMEM;
1230
1231 err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf);
1232 if (err && err != -EUCLEAN) {
1233 ubi_err("error %d while reading %d bytes from PEB %d:%d, "
1234 "read %zd bytes", err, len, pnum, offset, read);
1235 goto error;
1236 }
1237
1238 err = check_pattern(buf, 0xFF, len);
1239 if (err == 0) {
1240 ubi_err("flash region at PEB %d:%d, length %d does not "
1241 "contain all 0xFF bytes", pnum, offset, len);
1242 goto fail;
1243 }
1244
1245 kfree(buf);
1246 return 0;
1247
1248fail:
1249 ubi_err("paranoid check failed for PEB %d", pnum);
1250 dbg_msg("hex dump of the %d-%d region", offset, offset + len);
1251 ubi_dbg_hexdump(buf, len);
1252 err = 1;
1253error:
1254 ubi_dbg_dump_stack();
1255 kfree(buf);
1256 return err;
1257}
1258
1259#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
diff --git a/drivers/mtd/ubi/kapi.c b/drivers/mtd/ubi/kapi.c
new file mode 100644
index 000000000000..d352c4575c3d
--- /dev/null
+++ b/drivers/mtd/ubi/kapi.c
@@ -0,0 +1,575 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/* This file mostly implements UBI kernel API functions */
22
23#include <linux/module.h>
24#include <linux/err.h>
25#include <asm/div64.h>
26#include "ubi.h"
27
28/**
29 * ubi_get_device_info - get information about UBI device.
30 * @ubi_num: UBI device number
31 * @di: the information is stored here
32 *
33 * This function returns %0 in case of success and a %-ENODEV if there is no
34 * such UBI device.
35 */
36int ubi_get_device_info(int ubi_num, struct ubi_device_info *di)
37{
38 const struct ubi_device *ubi;
39
40 if (!try_module_get(THIS_MODULE))
41 return -ENODEV;
42
43 if (ubi_num < 0 || ubi_num >= UBI_MAX_DEVICES ||
44 !ubi_devices[ubi_num]) {
45 module_put(THIS_MODULE);
46 return -ENODEV;
47 }
48
49 ubi = ubi_devices[ubi_num];
50 di->ubi_num = ubi->ubi_num;
51 di->leb_size = ubi->leb_size;
52 di->min_io_size = ubi->min_io_size;
53 di->ro_mode = ubi->ro_mode;
54 di->cdev = MKDEV(ubi->major, 0);
55 module_put(THIS_MODULE);
56 return 0;
57}
58EXPORT_SYMBOL_GPL(ubi_get_device_info);
59
60/**
61 * ubi_get_volume_info - get information about UBI volume.
62 * @desc: volume descriptor
63 * @vi: the information is stored here
64 */
65void ubi_get_volume_info(struct ubi_volume_desc *desc,
66 struct ubi_volume_info *vi)
67{
68 const struct ubi_volume *vol = desc->vol;
69 const struct ubi_device *ubi = vol->ubi;
70
71 vi->vol_id = vol->vol_id;
72 vi->ubi_num = ubi->ubi_num;
73 vi->size = vol->reserved_pebs;
74 vi->used_bytes = vol->used_bytes;
75 vi->vol_type = vol->vol_type;
76 vi->corrupted = vol->corrupted;
77 vi->upd_marker = vol->upd_marker;
78 vi->alignment = vol->alignment;
79 vi->usable_leb_size = vol->usable_leb_size;
80 vi->name_len = vol->name_len;
81 vi->name = vol->name;
82 vi->cdev = MKDEV(ubi->major, vi->vol_id + 1);
83}
84EXPORT_SYMBOL_GPL(ubi_get_volume_info);
85
86/**
87 * ubi_open_volume - open UBI volume.
88 * @ubi_num: UBI device number
89 * @vol_id: volume ID
90 * @mode: open mode
91 *
92 * The @mode parameter specifies if the volume should be opened in read-only
93 * mode, read-write mode, or exclusive mode. The exclusive mode guarantees that
94 * nobody else will be able to open this volume. UBI allows to have many volume
95 * readers and one writer at a time.
96 *
97 * If a static volume is being opened for the first time since boot, it will be
98 * checked by this function, which means it will be fully read and the CRC
99 * checksum of each logical eraseblock will be checked.
100 *
101 * This function returns volume descriptor in case of success and a negative
102 * error code in case of failure.
103 */
104struct ubi_volume_desc *ubi_open_volume(int ubi_num, int vol_id, int mode)
105{
106 int err;
107 struct ubi_volume_desc *desc;
108 struct ubi_device *ubi = ubi_devices[ubi_num];
109 struct ubi_volume *vol;
110
111 dbg_msg("open device %d volume %d, mode %d", ubi_num, vol_id, mode);
112
113 err = -ENODEV;
114 if (!try_module_get(THIS_MODULE))
115 return ERR_PTR(err);
116
117 if (ubi_num < 0 || ubi_num >= UBI_MAX_DEVICES || !ubi)
118 goto out_put;
119
120 err = -EINVAL;
121 if (vol_id < 0 || vol_id >= ubi->vtbl_slots)
122 goto out_put;
123 if (mode != UBI_READONLY && mode != UBI_READWRITE &&
124 mode != UBI_EXCLUSIVE)
125 goto out_put;
126
127 desc = kmalloc(sizeof(struct ubi_volume_desc), GFP_KERNEL);
128 if (!desc) {
129 err = -ENOMEM;
130 goto out_put;
131 }
132
133 spin_lock(&ubi->volumes_lock);
134 vol = ubi->volumes[vol_id];
135 if (!vol) {
136 err = -ENODEV;
137 goto out_unlock;
138 }
139
140 err = -EBUSY;
141 switch (mode) {
142 case UBI_READONLY:
143 if (vol->exclusive)
144 goto out_unlock;
145 vol->readers += 1;
146 break;
147
148 case UBI_READWRITE:
149 if (vol->exclusive || vol->writers > 0)
150 goto out_unlock;
151 vol->writers += 1;
152 break;
153
154 case UBI_EXCLUSIVE:
155 if (vol->exclusive || vol->writers || vol->readers)
156 goto out_unlock;
157 vol->exclusive = 1;
158 break;
159 }
160 spin_unlock(&ubi->volumes_lock);
161
162 desc->vol = vol;
163 desc->mode = mode;
164
165 /*
166 * To prevent simultaneous checks of the same volume we use @vtbl_mutex,
167 * although it is not the purpose it was introduced for.
168 */
169 mutex_lock(&ubi->vtbl_mutex);
170 if (!vol->checked) {
171 /* This is the first open - check the volume */
172 err = ubi_check_volume(ubi, vol_id);
173 if (err < 0) {
174 mutex_unlock(&ubi->vtbl_mutex);
175 ubi_close_volume(desc);
176 return ERR_PTR(err);
177 }
178 if (err == 1) {
179 ubi_warn("volume %d on UBI device %d is corrupted",
180 vol_id, ubi->ubi_num);
181 vol->corrupted = 1;
182 }
183 vol->checked = 1;
184 }
185 mutex_unlock(&ubi->vtbl_mutex);
186 return desc;
187
188out_unlock:
189 spin_unlock(&ubi->volumes_lock);
190 kfree(desc);
191out_put:
192 module_put(THIS_MODULE);
193 return ERR_PTR(err);
194}
195EXPORT_SYMBOL_GPL(ubi_open_volume);
196
197/**
198 * ubi_open_volume_nm - open UBI volume by name.
199 * @ubi_num: UBI device number
200 * @name: volume name
201 * @mode: open mode
202 *
203 * This function is similar to 'ubi_open_volume()', but opens a volume by name.
204 */
205struct ubi_volume_desc *ubi_open_volume_nm(int ubi_num, const char *name,
206 int mode)
207{
208 int i, vol_id = -1, len;
209 struct ubi_volume_desc *ret;
210 struct ubi_device *ubi;
211
212 dbg_msg("open volume %s, mode %d", name, mode);
213
214 if (!name)
215 return ERR_PTR(-EINVAL);
216
217 len = strnlen(name, UBI_VOL_NAME_MAX + 1);
218 if (len > UBI_VOL_NAME_MAX)
219 return ERR_PTR(-EINVAL);
220
221 ret = ERR_PTR(-ENODEV);
222 if (!try_module_get(THIS_MODULE))
223 return ret;
224
225 if (ubi_num < 0 || ubi_num >= UBI_MAX_DEVICES || !ubi_devices[ubi_num])
226 goto out_put;
227
228 ubi = ubi_devices[ubi_num];
229
230 spin_lock(&ubi->volumes_lock);
231 /* Walk all volumes of this UBI device */
232 for (i = 0; i < ubi->vtbl_slots; i++) {
233 struct ubi_volume *vol = ubi->volumes[i];
234
235 if (vol && len == vol->name_len && !strcmp(name, vol->name)) {
236 vol_id = i;
237 break;
238 }
239 }
240 spin_unlock(&ubi->volumes_lock);
241
242 if (vol_id < 0)
243 goto out_put;
244
245 ret = ubi_open_volume(ubi_num, vol_id, mode);
246
247out_put:
248 module_put(THIS_MODULE);
249 return ret;
250}
251EXPORT_SYMBOL_GPL(ubi_open_volume_nm);
252
253/**
254 * ubi_close_volume - close UBI volume.
255 * @desc: volume descriptor
256 */
257void ubi_close_volume(struct ubi_volume_desc *desc)
258{
259 struct ubi_volume *vol = desc->vol;
260
261 dbg_msg("close volume %d, mode %d", vol->vol_id, desc->mode);
262
263 spin_lock(&vol->ubi->volumes_lock);
264 switch (desc->mode) {
265 case UBI_READONLY:
266 vol->readers -= 1;
267 break;
268 case UBI_READWRITE:
269 vol->writers -= 1;
270 break;
271 case UBI_EXCLUSIVE:
272 vol->exclusive = 0;
273 }
274 spin_unlock(&vol->ubi->volumes_lock);
275
276 kfree(desc);
277 module_put(THIS_MODULE);
278}
279EXPORT_SYMBOL_GPL(ubi_close_volume);
280
281/**
282 * ubi_leb_read - read data.
283 * @desc: volume descriptor
284 * @lnum: logical eraseblock number to read from
285 * @buf: buffer where to store the read data
286 * @offset: offset within the logical eraseblock to read from
287 * @len: how many bytes to read
288 * @check: whether UBI has to check the read data's CRC or not.
289 *
290 * This function reads data from offset @offset of logical eraseblock @lnum and
291 * stores the data at @buf. When reading from static volumes, @check specifies
292 * whether the data has to be checked or not. If yes, the whole logical
293 * eraseblock will be read and its CRC checksum will be checked (i.e., the CRC
294 * checksum is per-eraseblock). So checking may substantially slow down the
295 * read speed. The @check argument is ignored for dynamic volumes.
296 *
297 * In case of success, this function returns zero. In case of failure, this
298 * function returns a negative error code.
299 *
300 * %-EBADMSG error code is returned:
301 * o for both static and dynamic volumes if MTD driver has detected a data
302 * integrity problem (unrecoverable ECC checksum mismatch in case of NAND);
303 * o for static volumes in case of data CRC mismatch.
304 *
305 * If the volume is damaged because of an interrupted update this function just
306 * returns immediately with %-EBADF error code.
307 */
308int ubi_leb_read(struct ubi_volume_desc *desc, int lnum, char *buf, int offset,
309 int len, int check)
310{
311 struct ubi_volume *vol = desc->vol;
312 struct ubi_device *ubi = vol->ubi;
313 int err, vol_id = vol->vol_id;
314
315 dbg_msg("read %d bytes from LEB %d:%d:%d", len, vol_id, lnum, offset);
316
317 if (vol_id < 0 || vol_id >= ubi->vtbl_slots || lnum < 0 ||
318 lnum >= vol->used_ebs || offset < 0 || len < 0 ||
319 offset + len > vol->usable_leb_size)
320 return -EINVAL;
321
322 if (vol->vol_type == UBI_STATIC_VOLUME && lnum == vol->used_ebs - 1 &&
323 offset + len > vol->last_eb_bytes)
324 return -EINVAL;
325
326 if (vol->upd_marker)
327 return -EBADF;
328 if (len == 0)
329 return 0;
330
331 err = ubi_eba_read_leb(ubi, vol_id, lnum, buf, offset, len, check);
332 if (err && err == -EBADMSG && vol->vol_type == UBI_STATIC_VOLUME) {
333 ubi_warn("mark volume %d as corrupted", vol_id);
334 vol->corrupted = 1;
335 }
336
337 return err;
338}
339EXPORT_SYMBOL_GPL(ubi_leb_read);
340
341/**
342 * ubi_leb_write - write data.
343 * @desc: volume descriptor
344 * @lnum: logical eraseblock number to write to
345 * @buf: data to write
346 * @offset: offset within the logical eraseblock where to write
347 * @len: how many bytes to write
348 * @dtype: expected data type
349 *
350 * This function writes @len bytes of data from @buf to offset @offset of
351 * logical eraseblock @lnum. The @dtype argument describes expected lifetime of
352 * the data.
353 *
354 * This function takes care of physical eraseblock write failures. If write to
355 * the physical eraseblock write operation fails, the logical eraseblock is
356 * re-mapped to another physical eraseblock, the data is recovered, and the
357 * write finishes. UBI has a pool of reserved physical eraseblocks for this.
358 *
359 * If all the data were successfully written, zero is returned. If an error
360 * occurred and UBI has not been able to recover from it, this function returns
361 * a negative error code. Note, in case of an error, it is possible that
362 * something was still written to the flash media, but that may be some
363 * garbage.
364 *
365 * If the volume is damaged because of an interrupted update this function just
366 * returns immediately with %-EBADF code.
367 */
368int ubi_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf,
369 int offset, int len, int dtype)
370{
371 struct ubi_volume *vol = desc->vol;
372 struct ubi_device *ubi = vol->ubi;
373 int vol_id = vol->vol_id;
374
375 dbg_msg("write %d bytes to LEB %d:%d:%d", len, vol_id, lnum, offset);
376
377 if (vol_id < 0 || vol_id >= ubi->vtbl_slots)
378 return -EINVAL;
379
380 if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
381 return -EROFS;
382
383 if (lnum < 0 || lnum >= vol->reserved_pebs || offset < 0 || len < 0 ||
384 offset + len > vol->usable_leb_size || offset % ubi->min_io_size ||
385 len % ubi->min_io_size)
386 return -EINVAL;
387
388 if (dtype != UBI_LONGTERM && dtype != UBI_SHORTTERM &&
389 dtype != UBI_UNKNOWN)
390 return -EINVAL;
391
392 if (vol->upd_marker)
393 return -EBADF;
394
395 if (len == 0)
396 return 0;
397
398 return ubi_eba_write_leb(ubi, vol_id, lnum, buf, offset, len, dtype);
399}
400EXPORT_SYMBOL_GPL(ubi_leb_write);
401
402/*
403 * ubi_leb_change - change logical eraseblock atomically.
404 * @desc: volume descriptor
405 * @lnum: logical eraseblock number to change
406 * @buf: data to write
407 * @len: how many bytes to write
408 * @dtype: expected data type
409 *
410 * This function changes the contents of a logical eraseblock atomically. @buf
411 * has to contain new logical eraseblock data, and @len - the length of the
412 * data, which has to be aligned. The length may be shorter then the logical
413 * eraseblock size, ant the logical eraseblock may be appended to more times
414 * later on. This function guarantees that in case of an unclean reboot the old
415 * contents is preserved. Returns zero in case of success and a negative error
416 * code in case of failure.
417 */
418int ubi_leb_change(struct ubi_volume_desc *desc, int lnum, const void *buf,
419 int len, int dtype)
420{
421 struct ubi_volume *vol = desc->vol;
422 struct ubi_device *ubi = vol->ubi;
423 int vol_id = vol->vol_id;
424
425 dbg_msg("atomically write %d bytes to LEB %d:%d", len, vol_id, lnum);
426
427 if (vol_id < 0 || vol_id >= ubi->vtbl_slots)
428 return -EINVAL;
429
430 if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
431 return -EROFS;
432
433 if (lnum < 0 || lnum >= vol->reserved_pebs || len < 0 ||
434 len > vol->usable_leb_size || len % ubi->min_io_size)
435 return -EINVAL;
436
437 if (dtype != UBI_LONGTERM && dtype != UBI_SHORTTERM &&
438 dtype != UBI_UNKNOWN)
439 return -EINVAL;
440
441 if (vol->upd_marker)
442 return -EBADF;
443
444 if (len == 0)
445 return 0;
446
447 return ubi_eba_atomic_leb_change(ubi, vol_id, lnum, buf, len, dtype);
448}
449EXPORT_SYMBOL_GPL(ubi_leb_change);
450
451/**
452 * ubi_leb_erase - erase logical eraseblock.
453 * @desc: volume descriptor
454 * @lnum: logical eraseblock number
455 *
456 * This function un-maps logical eraseblock @lnum and synchronously erases the
457 * correspondent physical eraseblock. Returns zero in case of success and a
458 * negative error code in case of failure.
459 *
460 * If the volume is damaged because of an interrupted update this function just
461 * returns immediately with %-EBADF code.
462 */
463int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum)
464{
465 struct ubi_volume *vol = desc->vol;
466 struct ubi_device *ubi = vol->ubi;
467 int err, vol_id = vol->vol_id;
468
469 dbg_msg("erase LEB %d:%d", vol_id, lnum);
470
471 if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
472 return -EROFS;
473
474 if (lnum < 0 || lnum >= vol->reserved_pebs)
475 return -EINVAL;
476
477 if (vol->upd_marker)
478 return -EBADF;
479
480 err = ubi_eba_unmap_leb(ubi, vol_id, lnum);
481 if (err)
482 return err;
483
484 return ubi_wl_flush(ubi);
485}
486EXPORT_SYMBOL_GPL(ubi_leb_erase);
487
488/**
489 * ubi_leb_unmap - un-map logical eraseblock.
490 * @desc: volume descriptor
491 * @lnum: logical eraseblock number
492 *
493 * This function un-maps logical eraseblock @lnum and schedules the
494 * corresponding physical eraseblock for erasure, so that it will eventually be
495 * physically erased in background. This operation is much faster then the
496 * erase operation.
497 *
498 * Unlike erase, the un-map operation does not guarantee that the logical
499 * eraseblock will contain all 0xFF bytes when UBI is initialized again. For
500 * example, if several logical eraseblocks are un-mapped, and an unclean reboot
501 * happens after this, the logical eraseblocks will not necessarily be
502 * un-mapped again when this MTD device is attached. They may actually be
503 * mapped to the same physical eraseblocks again. So, this function has to be
504 * used with care.
505 *
506 * In other words, when un-mapping a logical eraseblock, UBI does not store
507 * any information about this on the flash media, it just marks the logical
508 * eraseblock as "un-mapped" in RAM. If UBI is detached before the physical
509 * eraseblock is physically erased, it will be mapped again to the same logical
510 * eraseblock when the MTD device is attached again.
511 *
512 * The main and obvious use-case of this function is when the contents of a
513 * logical eraseblock has to be re-written. Then it is much more efficient to
514 * first un-map it, then write new data, rather then first erase it, then write
515 * new data. Note, once new data has been written to the logical eraseblock,
516 * UBI guarantees that the old contents has gone forever. In other words, if an
517 * unclean reboot happens after the logical eraseblock has been un-mapped and
518 * then written to, it will contain the last written data.
519 *
520 * This function returns zero in case of success and a negative error code in
521 * case of failure. If the volume is damaged because of an interrupted update
522 * this function just returns immediately with %-EBADF code.
523 */
524int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum)
525{
526 struct ubi_volume *vol = desc->vol;
527 struct ubi_device *ubi = vol->ubi;
528 int vol_id = vol->vol_id;
529
530 dbg_msg("unmap LEB %d:%d", vol_id, lnum);
531
532 if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
533 return -EROFS;
534
535 if (lnum < 0 || lnum >= vol->reserved_pebs)
536 return -EINVAL;
537
538 if (vol->upd_marker)
539 return -EBADF;
540
541 return ubi_eba_unmap_leb(ubi, vol_id, lnum);
542}
543EXPORT_SYMBOL_GPL(ubi_leb_unmap);
544
545/**
546 * ubi_is_mapped - check if logical eraseblock is mapped.
547 * @desc: volume descriptor
548 * @lnum: logical eraseblock number
549 *
550 * This function checks if logical eraseblock @lnum is mapped to a physical
551 * eraseblock. If a logical eraseblock is un-mapped, this does not necessarily
552 * mean it will still be un-mapped after the UBI device is re-attached. The
553 * logical eraseblock may become mapped to the physical eraseblock it was last
554 * mapped to.
555 *
556 * This function returns %1 if the LEB is mapped, %0 if not, and a negative
557 * error code in case of failure. If the volume is damaged because of an
558 * interrupted update this function just returns immediately with %-EBADF error
559 * code.
560 */
561int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum)
562{
563 struct ubi_volume *vol = desc->vol;
564
565 dbg_msg("test LEB %d:%d", vol->vol_id, lnum);
566
567 if (lnum < 0 || lnum >= vol->reserved_pebs)
568 return -EINVAL;
569
570 if (vol->upd_marker)
571 return -EBADF;
572
573 return vol->eba_tbl[lnum] >= 0;
574}
575EXPORT_SYMBOL_GPL(ubi_is_mapped);
diff --git a/drivers/mtd/ubi/misc.c b/drivers/mtd/ubi/misc.c
new file mode 100644
index 000000000000..38d4e6757dc7
--- /dev/null
+++ b/drivers/mtd/ubi/misc.c
@@ -0,0 +1,105 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/* Here we keep miscellaneous functions which are used all over the UBI code */
22
23#include "ubi.h"
24
25/**
26 * calc_data_len - calculate how much real data is stored in a buffer.
27 * @ubi: UBI device description object
28 * @buf: a buffer with the contents of the physical eraseblock
29 * @length: the buffer length
30 *
31 * This function calculates how much "real data" is stored in @buf and returnes
32 * the length. Continuous 0xFF bytes at the end of the buffer are not
33 * considered as "real data".
34 */
35int ubi_calc_data_len(const struct ubi_device *ubi, const void *buf,
36 int length)
37{
38 int i;
39
40 ubi_assert(length % ubi->min_io_size == 0);
41
42 for (i = length - 1; i >= 0; i--)
43 if (((const uint8_t *)buf)[i] != 0xFF)
44 break;
45
46 /* The resulting length must be aligned to the minimum flash I/O size */
47 length = ALIGN(i + 1, ubi->min_io_size);
48 return length;
49}
50
51/**
52 * ubi_check_volume - check the contents of a static volume.
53 * @ubi: UBI device description object
54 * @vol_id: ID of the volume to check
55 *
56 * This function checks if static volume @vol_id is corrupted by fully reading
57 * it and checking data CRC. This function returns %0 if the volume is not
58 * corrupted, %1 if it is corrupted and a negative error code in case of
59 * failure. Dynamic volumes are not checked and zero is returned immediately.
60 */
61int ubi_check_volume(struct ubi_device *ubi, int vol_id)
62{
63 void *buf;
64 int err = 0, i;
65 struct ubi_volume *vol = ubi->volumes[vol_id];
66
67 if (vol->vol_type != UBI_STATIC_VOLUME)
68 return 0;
69
70 buf = kmalloc(vol->usable_leb_size, GFP_KERNEL);
71 if (!buf)
72 return -ENOMEM;
73
74 for (i = 0; i < vol->used_ebs; i++) {
75 int size;
76
77 if (i == vol->used_ebs - 1)
78 size = vol->last_eb_bytes;
79 else
80 size = vol->usable_leb_size;
81
82 err = ubi_eba_read_leb(ubi, vol_id, i, buf, 0, size, 1);
83 if (err) {
84 if (err == -EBADMSG)
85 err = 1;
86 break;
87 }
88 }
89
90 kfree(buf);
91 return err;
92}
93
94/**
95 * ubi_calculate_rsvd_pool - calculate how many PEBs must be reserved for bad
96 * eraseblock handling.
97 * @ubi: UBI device description object
98 */
99void ubi_calculate_reserved(struct ubi_device *ubi)
100{
101 ubi->beb_rsvd_level = ubi->good_peb_count/100;
102 ubi->beb_rsvd_level *= CONFIG_MTD_UBI_BEB_RESERVE;
103 if (ubi->beb_rsvd_level < MIN_RESEVED_PEBS)
104 ubi->beb_rsvd_level = MIN_RESEVED_PEBS;
105}
diff --git a/drivers/mtd/ubi/scan.c b/drivers/mtd/ubi/scan.c
new file mode 100644
index 000000000000..473f3200b868
--- /dev/null
+++ b/drivers/mtd/ubi/scan.c
@@ -0,0 +1,1368 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/*
22 * UBI scanning unit.
23 *
24 * This unit is responsible for scanning the flash media, checking UBI
25 * headers and providing complete information about the UBI flash image.
26 *
27 * The scanning information is reoresented by a &struct ubi_scan_info' object.
28 * Information about found volumes is represented by &struct ubi_scan_volume
29 * objects which are kept in volume RB-tree with root at the @volumes field.
30 * The RB-tree is indexed by the volume ID.
31 *
32 * Found logical eraseblocks are represented by &struct ubi_scan_leb objects.
33 * These objects are kept in per-volume RB-trees with the root at the
34 * corresponding &struct ubi_scan_volume object. To put it differently, we keep
35 * an RB-tree of per-volume objects and each of these objects is the root of
36 * RB-tree of per-eraseblock objects.
37 *
38 * Corrupted physical eraseblocks are put to the @corr list, free physical
39 * eraseblocks are put to the @free list and the physical eraseblock to be
40 * erased are put to the @erase list.
41 */
42
43#include <linux/err.h>
44#include <linux/crc32.h>
45#include "ubi.h"
46
47#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
48static int paranoid_check_si(const struct ubi_device *ubi,
49 struct ubi_scan_info *si);
50#else
51#define paranoid_check_si(ubi, si) 0
52#endif
53
54/* Temporary variables used during scanning */
55static struct ubi_ec_hdr *ech;
56static struct ubi_vid_hdr *vidh;
57
58int ubi_scan_add_to_list(struct ubi_scan_info *si, int pnum, int ec,
59 struct list_head *list)
60{
61 struct ubi_scan_leb *seb;
62
63 if (list == &si->free)
64 dbg_bld("add to free: PEB %d, EC %d", pnum, ec);
65 else if (list == &si->erase)
66 dbg_bld("add to erase: PEB %d, EC %d", pnum, ec);
67 else if (list == &si->corr)
68 dbg_bld("add to corrupted: PEB %d, EC %d", pnum, ec);
69 else if (list == &si->alien)
70 dbg_bld("add to alien: PEB %d, EC %d", pnum, ec);
71 else
72 BUG();
73
74 seb = kmalloc(sizeof(struct ubi_scan_leb), GFP_KERNEL);
75 if (!seb)
76 return -ENOMEM;
77
78 seb->pnum = pnum;
79 seb->ec = ec;
80 list_add_tail(&seb->u.list, list);
81 return 0;
82}
83
84/**
85 * commit_to_mean_value - commit intermediate results to the final mean erase
86 * counter value.
87 * @si: scanning information
88 *
89 * This is a helper function which calculates partial mean erase counter mean
90 * value and adds it to the resulting mean value. As we can work only in
91 * integer arithmetic and we want to calculate the mean value of erase counter
92 * accurately, we first sum erase counter values in @si->ec_sum variable and
93 * count these components in @si->ec_count. If this temporary @si->ec_sum is
94 * going to overflow, we calculate the partial mean value
95 * (@si->ec_sum/@si->ec_count) and add it to @si->mean_ec.
96 */
97static void commit_to_mean_value(struct ubi_scan_info *si)
98{
99 si->ec_sum /= si->ec_count;
100 if (si->ec_sum % si->ec_count >= si->ec_count / 2)
101 si->mean_ec += 1;
102 si->mean_ec += si->ec_sum;
103}
104
105/**
106 * validate_vid_hdr - check that volume identifier header is correct and
107 * consistent.
108 * @vid_hdr: the volume identifier header to check
109 * @sv: information about the volume this logical eraseblock belongs to
110 * @pnum: physical eraseblock number the VID header came from
111 *
112 * This function checks that data stored in @vid_hdr is consistent. Returns
113 * non-zero if an inconsistency was found and zero if not.
114 *
115 * Note, UBI does sanity check of everything it reads from the flash media.
116 * Most of the checks are done in the I/O unit. Here we check that the
117 * information in the VID header is consistent to the information in other VID
118 * headers of the same volume.
119 */
120static int validate_vid_hdr(const struct ubi_vid_hdr *vid_hdr,
121 const struct ubi_scan_volume *sv, int pnum)
122{
123 int vol_type = vid_hdr->vol_type;
124 int vol_id = ubi32_to_cpu(vid_hdr->vol_id);
125 int used_ebs = ubi32_to_cpu(vid_hdr->used_ebs);
126 int data_pad = ubi32_to_cpu(vid_hdr->data_pad);
127
128 if (sv->leb_count != 0) {
129 int sv_vol_type;
130
131 /*
132 * This is not the first logical eraseblock belonging to this
133 * volume. Ensure that the data in its VID header is consistent
134 * to the data in previous logical eraseblock headers.
135 */
136
137 if (vol_id != sv->vol_id) {
138 dbg_err("inconsistent vol_id");
139 goto bad;
140 }
141
142 if (sv->vol_type == UBI_STATIC_VOLUME)
143 sv_vol_type = UBI_VID_STATIC;
144 else
145 sv_vol_type = UBI_VID_DYNAMIC;
146
147 if (vol_type != sv_vol_type) {
148 dbg_err("inconsistent vol_type");
149 goto bad;
150 }
151
152 if (used_ebs != sv->used_ebs) {
153 dbg_err("inconsistent used_ebs");
154 goto bad;
155 }
156
157 if (data_pad != sv->data_pad) {
158 dbg_err("inconsistent data_pad");
159 goto bad;
160 }
161 }
162
163 return 0;
164
165bad:
166 ubi_err("inconsistent VID header at PEB %d", pnum);
167 ubi_dbg_dump_vid_hdr(vid_hdr);
168 ubi_dbg_dump_sv(sv);
169 return -EINVAL;
170}
171
172/**
173 * add_volume - add volume to the scanning information.
174 * @si: scanning information
175 * @vol_id: ID of the volume to add
176 * @pnum: physical eraseblock number
177 * @vid_hdr: volume identifier header
178 *
179 * If the volume corresponding to the @vid_hdr logical eraseblock is already
180 * present in the scanning information, this function does nothing. Otherwise
181 * it adds corresponding volume to the scanning information. Returns a pointer
182 * to the scanning volume object in case of success and a negative error code
183 * in case of failure.
184 */
185static struct ubi_scan_volume *add_volume(struct ubi_scan_info *si, int vol_id,
186 int pnum,
187 const struct ubi_vid_hdr *vid_hdr)
188{
189 struct ubi_scan_volume *sv;
190 struct rb_node **p = &si->volumes.rb_node, *parent = NULL;
191
192 ubi_assert(vol_id == ubi32_to_cpu(vid_hdr->vol_id));
193
194 /* Walk the volume RB-tree to look if this volume is already present */
195 while (*p) {
196 parent = *p;
197 sv = rb_entry(parent, struct ubi_scan_volume, rb);
198
199 if (vol_id == sv->vol_id)
200 return sv;
201
202 if (vol_id > sv->vol_id)
203 p = &(*p)->rb_left;
204 else
205 p = &(*p)->rb_right;
206 }
207
208 /* The volume is absent - add it */
209 sv = kmalloc(sizeof(struct ubi_scan_volume), GFP_KERNEL);
210 if (!sv)
211 return ERR_PTR(-ENOMEM);
212
213 sv->highest_lnum = sv->leb_count = 0;
214 si->max_sqnum = 0;
215 sv->vol_id = vol_id;
216 sv->root = RB_ROOT;
217 sv->used_ebs = ubi32_to_cpu(vid_hdr->used_ebs);
218 sv->data_pad = ubi32_to_cpu(vid_hdr->data_pad);
219 sv->compat = vid_hdr->compat;
220 sv->vol_type = vid_hdr->vol_type == UBI_VID_DYNAMIC ? UBI_DYNAMIC_VOLUME
221 : UBI_STATIC_VOLUME;
222 if (vol_id > si->highest_vol_id)
223 si->highest_vol_id = vol_id;
224
225 rb_link_node(&sv->rb, parent, p);
226 rb_insert_color(&sv->rb, &si->volumes);
227 si->vols_found += 1;
228 dbg_bld("added volume %d", vol_id);
229 return sv;
230}
231
232/**
233 * compare_lebs - find out which logical eraseblock is newer.
234 * @ubi: UBI device description object
235 * @seb: first logical eraseblock to compare
236 * @pnum: physical eraseblock number of the second logical eraseblock to
237 * compare
238 * @vid_hdr: volume identifier header of the second logical eraseblock
239 *
240 * This function compares 2 copies of a LEB and informs which one is newer. In
241 * case of success this function returns a positive value, in case of failure, a
242 * negative error code is returned. The success return codes use the following
243 * bits:
244 * o bit 0 is cleared: the first PEB (described by @seb) is newer then the
245 * second PEB (described by @pnum and @vid_hdr);
246 * o bit 0 is set: the second PEB is newer;
247 * o bit 1 is cleared: no bit-flips were detected in the newer LEB;
248 * o bit 1 is set: bit-flips were detected in the newer LEB;
249 * o bit 2 is cleared: the older LEB is not corrupted;
250 * o bit 2 is set: the older LEB is corrupted.
251 */
252static int compare_lebs(const struct ubi_device *ubi,
253 const struct ubi_scan_leb *seb, int pnum,
254 const struct ubi_vid_hdr *vid_hdr)
255{
256 void *buf;
257 int len, err, second_is_newer, bitflips = 0, corrupted = 0;
258 uint32_t data_crc, crc;
259 struct ubi_vid_hdr *vidh = NULL;
260 unsigned long long sqnum2 = ubi64_to_cpu(vid_hdr->sqnum);
261
262 if (seb->sqnum == 0 && sqnum2 == 0) {
263 long long abs, v1 = seb->leb_ver, v2 = ubi32_to_cpu(vid_hdr->leb_ver);
264
265 /*
266 * UBI constantly increases the logical eraseblock version
267 * number and it can overflow. Thus, we have to bear in mind
268 * that versions that are close to %0xFFFFFFFF are less then
269 * versions that are close to %0.
270 *
271 * The UBI WL unit guarantees that the number of pending tasks
272 * is not greater then %0x7FFFFFFF. So, if the difference
273 * between any two versions is greater or equivalent to
274 * %0x7FFFFFFF, there was an overflow and the logical
275 * eraseblock with lower version is actually newer then the one
276 * with higher version.
277 *
278 * FIXME: but this is anyway obsolete and will be removed at
279 * some point.
280 */
281
282 dbg_bld("using old crappy leb_ver stuff");
283
284 abs = v1 - v2;
285 if (abs < 0)
286 abs = -abs;
287
288 if (abs < 0x7FFFFFFF)
289 /* Non-overflow situation */
290 second_is_newer = (v2 > v1);
291 else
292 second_is_newer = (v2 < v1);
293 } else
294 /* Obviously the LEB with lower sequence counter is older */
295 second_is_newer = sqnum2 > seb->sqnum;
296
297 /*
298 * Now we know which copy is newer. If the copy flag of the PEB with
299 * newer version is not set, then we just return, otherwise we have to
300 * check data CRC. For the second PEB we already have the VID header,
301 * for the first one - we'll need to re-read it from flash.
302 *
303 * FIXME: this may be optimized so that we wouldn't read twice.
304 */
305
306 if (second_is_newer) {
307 if (!vid_hdr->copy_flag) {
308 /* It is not a copy, so it is newer */
309 dbg_bld("second PEB %d is newer, copy_flag is unset",
310 pnum);
311 return 1;
312 }
313 } else {
314 pnum = seb->pnum;
315
316 vidh = ubi_zalloc_vid_hdr(ubi);
317 if (!vidh)
318 return -ENOMEM;
319
320 err = ubi_io_read_vid_hdr(ubi, pnum, vidh, 0);
321 if (err) {
322 if (err == UBI_IO_BITFLIPS)
323 bitflips = 1;
324 else {
325 dbg_err("VID of PEB %d header is bad, but it "
326 "was OK earlier", pnum);
327 if (err > 0)
328 err = -EIO;
329
330 goto out_free_vidh;
331 }
332 }
333
334 if (!vidh->copy_flag) {
335 /* It is not a copy, so it is newer */
336 dbg_bld("first PEB %d is newer, copy_flag is unset",
337 pnum);
338 err = bitflips << 1;
339 goto out_free_vidh;
340 }
341
342 vid_hdr = vidh;
343 }
344
345 /* Read the data of the copy and check the CRC */
346
347 len = ubi32_to_cpu(vid_hdr->data_size);
348 buf = kmalloc(len, GFP_KERNEL);
349 if (!buf) {
350 err = -ENOMEM;
351 goto out_free_vidh;
352 }
353
354 err = ubi_io_read_data(ubi, buf, pnum, 0, len);
355 if (err && err != UBI_IO_BITFLIPS)
356 goto out_free_buf;
357
358 data_crc = ubi32_to_cpu(vid_hdr->data_crc);
359 crc = crc32(UBI_CRC32_INIT, buf, len);
360 if (crc != data_crc) {
361 dbg_bld("PEB %d CRC error: calculated %#08x, must be %#08x",
362 pnum, crc, data_crc);
363 corrupted = 1;
364 bitflips = 0;
365 second_is_newer = !second_is_newer;
366 } else {
367 dbg_bld("PEB %d CRC is OK", pnum);
368 bitflips = !!err;
369 }
370
371 kfree(buf);
372 ubi_free_vid_hdr(ubi, vidh);
373
374 if (second_is_newer)
375 dbg_bld("second PEB %d is newer, copy_flag is set", pnum);
376 else
377 dbg_bld("first PEB %d is newer, copy_flag is set", pnum);
378
379 return second_is_newer | (bitflips << 1) | (corrupted << 2);
380
381out_free_buf:
382 kfree(buf);
383out_free_vidh:
384 ubi_free_vid_hdr(ubi, vidh);
385 ubi_assert(err < 0);
386 return err;
387}
388
389/**
390 * ubi_scan_add_used - add information about a physical eraseblock to the
391 * scanning information.
392 * @ubi: UBI device description object
393 * @si: scanning information
394 * @pnum: the physical eraseblock number
395 * @ec: erase counter
396 * @vid_hdr: the volume identifier header
397 * @bitflips: if bit-flips were detected when this physical eraseblock was read
398 *
399 * This function returns zero in case of success and a negative error code in
400 * case of failure.
401 */
402int ubi_scan_add_used(const struct ubi_device *ubi, struct ubi_scan_info *si,
403 int pnum, int ec, const struct ubi_vid_hdr *vid_hdr,
404 int bitflips)
405{
406 int err, vol_id, lnum;
407 uint32_t leb_ver;
408 unsigned long long sqnum;
409 struct ubi_scan_volume *sv;
410 struct ubi_scan_leb *seb;
411 struct rb_node **p, *parent = NULL;
412
413 vol_id = ubi32_to_cpu(vid_hdr->vol_id);
414 lnum = ubi32_to_cpu(vid_hdr->lnum);
415 sqnum = ubi64_to_cpu(vid_hdr->sqnum);
416 leb_ver = ubi32_to_cpu(vid_hdr->leb_ver);
417
418 dbg_bld("PEB %d, LEB %d:%d, EC %d, sqnum %llu, ver %u, bitflips %d",
419 pnum, vol_id, lnum, ec, sqnum, leb_ver, bitflips);
420
421 sv = add_volume(si, vol_id, pnum, vid_hdr);
422 if (IS_ERR(sv) < 0)
423 return PTR_ERR(sv);
424
425 /*
426 * Walk the RB-tree of logical eraseblocks of volume @vol_id to look
427 * if this is the first instance of this logical eraseblock or not.
428 */
429 p = &sv->root.rb_node;
430 while (*p) {
431 int cmp_res;
432
433 parent = *p;
434 seb = rb_entry(parent, struct ubi_scan_leb, u.rb);
435 if (lnum != seb->lnum) {
436 if (lnum < seb->lnum)
437 p = &(*p)->rb_left;
438 else
439 p = &(*p)->rb_right;
440 continue;
441 }
442
443 /*
444 * There is already a physical eraseblock describing the same
445 * logical eraseblock present.
446 */
447
448 dbg_bld("this LEB already exists: PEB %d, sqnum %llu, "
449 "LEB ver %u, EC %d", seb->pnum, seb->sqnum,
450 seb->leb_ver, seb->ec);
451
452 /*
453 * Make sure that the logical eraseblocks have different
454 * versions. Otherwise the image is bad.
455 */
456 if (seb->leb_ver == leb_ver && leb_ver != 0) {
457 ubi_err("two LEBs with same version %u", leb_ver);
458 ubi_dbg_dump_seb(seb, 0);
459 ubi_dbg_dump_vid_hdr(vid_hdr);
460 return -EINVAL;
461 }
462
463 /*
464 * Make sure that the logical eraseblocks have different
465 * sequence numbers. Otherwise the image is bad.
466 *
467 * FIXME: remove 'sqnum != 0' check when leb_ver is removed.
468 */
469 if (seb->sqnum == sqnum && sqnum != 0) {
470 ubi_err("two LEBs with same sequence number %llu",
471 sqnum);
472 ubi_dbg_dump_seb(seb, 0);
473 ubi_dbg_dump_vid_hdr(vid_hdr);
474 return -EINVAL;
475 }
476
477 /*
478 * Now we have to drop the older one and preserve the newer
479 * one.
480 */
481 cmp_res = compare_lebs(ubi, seb, pnum, vid_hdr);
482 if (cmp_res < 0)
483 return cmp_res;
484
485 if (cmp_res & 1) {
486 /*
487 * This logical eraseblock is newer then the one
488 * found earlier.
489 */
490 err = validate_vid_hdr(vid_hdr, sv, pnum);
491 if (err)
492 return err;
493
494 if (cmp_res & 4)
495 err = ubi_scan_add_to_list(si, seb->pnum,
496 seb->ec, &si->corr);
497 else
498 err = ubi_scan_add_to_list(si, seb->pnum,
499 seb->ec, &si->erase);
500 if (err)
501 return err;
502
503 seb->ec = ec;
504 seb->pnum = pnum;
505 seb->scrub = ((cmp_res & 2) || bitflips);
506 seb->sqnum = sqnum;
507 seb->leb_ver = leb_ver;
508
509 if (sv->highest_lnum == lnum)
510 sv->last_data_size =
511 ubi32_to_cpu(vid_hdr->data_size);
512
513 return 0;
514 } else {
515 /*
516 * This logical eraseblock is older then the one found
517 * previously.
518 */
519 if (cmp_res & 4)
520 return ubi_scan_add_to_list(si, pnum, ec,
521 &si->corr);
522 else
523 return ubi_scan_add_to_list(si, pnum, ec,
524 &si->erase);
525 }
526 }
527
528 /*
529 * We've met this logical eraseblock for the first time, add it to the
530 * scanning information.
531 */
532
533 err = validate_vid_hdr(vid_hdr, sv, pnum);
534 if (err)
535 return err;
536
537 seb = kmalloc(sizeof(struct ubi_scan_leb), GFP_KERNEL);
538 if (!seb)
539 return -ENOMEM;
540
541 seb->ec = ec;
542 seb->pnum = pnum;
543 seb->lnum = lnum;
544 seb->sqnum = sqnum;
545 seb->scrub = bitflips;
546 seb->leb_ver = leb_ver;
547
548 if (sv->highest_lnum <= lnum) {
549 sv->highest_lnum = lnum;
550 sv->last_data_size = ubi32_to_cpu(vid_hdr->data_size);
551 }
552
553 if (si->max_sqnum < sqnum)
554 si->max_sqnum = sqnum;
555
556 sv->leb_count += 1;
557 rb_link_node(&seb->u.rb, parent, p);
558 rb_insert_color(&seb->u.rb, &sv->root);
559 return 0;
560}
561
562/**
563 * ubi_scan_find_sv - find information about a particular volume in the
564 * scanning information.
565 * @si: scanning information
566 * @vol_id: the requested volume ID
567 *
568 * This function returns a pointer to the volume description or %NULL if there
569 * are no data about this volume in the scanning information.
570 */
571struct ubi_scan_volume *ubi_scan_find_sv(const struct ubi_scan_info *si,
572 int vol_id)
573{
574 struct ubi_scan_volume *sv;
575 struct rb_node *p = si->volumes.rb_node;
576
577 while (p) {
578 sv = rb_entry(p, struct ubi_scan_volume, rb);
579
580 if (vol_id == sv->vol_id)
581 return sv;
582
583 if (vol_id > sv->vol_id)
584 p = p->rb_left;
585 else
586 p = p->rb_right;
587 }
588
589 return NULL;
590}
591
592/**
593 * ubi_scan_find_seb - find information about a particular logical
594 * eraseblock in the volume scanning information.
595 * @sv: a pointer to the volume scanning information
596 * @lnum: the requested logical eraseblock
597 *
598 * This function returns a pointer to the scanning logical eraseblock or %NULL
599 * if there are no data about it in the scanning volume information.
600 */
601struct ubi_scan_leb *ubi_scan_find_seb(const struct ubi_scan_volume *sv,
602 int lnum)
603{
604 struct ubi_scan_leb *seb;
605 struct rb_node *p = sv->root.rb_node;
606
607 while (p) {
608 seb = rb_entry(p, struct ubi_scan_leb, u.rb);
609
610 if (lnum == seb->lnum)
611 return seb;
612
613 if (lnum > seb->lnum)
614 p = p->rb_left;
615 else
616 p = p->rb_right;
617 }
618
619 return NULL;
620}
621
622/**
623 * ubi_scan_rm_volume - delete scanning information about a volume.
624 * @si: scanning information
625 * @sv: the volume scanning information to delete
626 */
627void ubi_scan_rm_volume(struct ubi_scan_info *si, struct ubi_scan_volume *sv)
628{
629 struct rb_node *rb;
630 struct ubi_scan_leb *seb;
631
632 dbg_bld("remove scanning information about volume %d", sv->vol_id);
633
634 while ((rb = rb_first(&sv->root))) {
635 seb = rb_entry(rb, struct ubi_scan_leb, u.rb);
636 rb_erase(&seb->u.rb, &sv->root);
637 list_add_tail(&seb->u.list, &si->erase);
638 }
639
640 rb_erase(&sv->rb, &si->volumes);
641 kfree(sv);
642 si->vols_found -= 1;
643}
644
645/**
646 * ubi_scan_erase_peb - erase a physical eraseblock.
647 * @ubi: UBI device description object
648 * @si: scanning information
649 * @pnum: physical eraseblock number to erase;
650 * @ec: erase counter value to write (%UBI_SCAN_UNKNOWN_EC if it is unknown)
651 *
652 * This function erases physical eraseblock 'pnum', and writes the erase
653 * counter header to it. This function should only be used on UBI device
654 * initialization stages, when the EBA unit had not been yet initialized. This
655 * function returns zero in case of success and a negative error code in case
656 * of failure.
657 */
658int ubi_scan_erase_peb(const struct ubi_device *ubi,
659 const struct ubi_scan_info *si, int pnum, int ec)
660{
661 int err;
662 struct ubi_ec_hdr *ec_hdr;
663
664 ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
665 if (!ec_hdr)
666 return -ENOMEM;
667
668 if ((long long)ec >= UBI_MAX_ERASECOUNTER) {
669 /*
670 * Erase counter overflow. Upgrade UBI and use 64-bit
671 * erase counters internally.
672 */
673 ubi_err("erase counter overflow at PEB %d, EC %d", pnum, ec);
674 return -EINVAL;
675 }
676
677 ec_hdr->ec = cpu_to_ubi64(ec);
678
679 err = ubi_io_sync_erase(ubi, pnum, 0);
680 if (err < 0)
681 goto out_free;
682
683 err = ubi_io_write_ec_hdr(ubi, pnum, ec_hdr);
684
685out_free:
686 kfree(ec_hdr);
687 return err;
688}
689
690/**
691 * ubi_scan_get_free_peb - get a free physical eraseblock.
692 * @ubi: UBI device description object
693 * @si: scanning information
694 *
695 * This function returns a free physical eraseblock. It is supposed to be
696 * called on the UBI initialization stages when the wear-leveling unit is not
697 * initialized yet. This function picks a physical eraseblocks from one of the
698 * lists, writes the EC header if it is needed, and removes it from the list.
699 *
700 * This function returns scanning physical eraseblock information in case of
701 * success and an error code in case of failure.
702 */
703struct ubi_scan_leb *ubi_scan_get_free_peb(const struct ubi_device *ubi,
704 struct ubi_scan_info *si)
705{
706 int err = 0, i;
707 struct ubi_scan_leb *seb;
708
709 if (!list_empty(&si->free)) {
710 seb = list_entry(si->free.next, struct ubi_scan_leb, u.list);
711 list_del(&seb->u.list);
712 dbg_bld("return free PEB %d, EC %d", seb->pnum, seb->ec);
713 return seb;
714 }
715
716 for (i = 0; i < 2; i++) {
717 struct list_head *head;
718 struct ubi_scan_leb *tmp_seb;
719
720 if (i == 0)
721 head = &si->erase;
722 else
723 head = &si->corr;
724
725 /*
726 * We try to erase the first physical eraseblock from the @head
727 * list and pick it if we succeed, or try to erase the
728 * next one if not. And so forth. We don't want to take care
729 * about bad eraseblocks here - they'll be handled later.
730 */
731 list_for_each_entry_safe(seb, tmp_seb, head, u.list) {
732 if (seb->ec == UBI_SCAN_UNKNOWN_EC)
733 seb->ec = si->mean_ec;
734
735 err = ubi_scan_erase_peb(ubi, si, seb->pnum, seb->ec+1);
736 if (err)
737 continue;
738
739 seb->ec += 1;
740 list_del(&seb->u.list);
741 dbg_bld("return PEB %d, EC %d", seb->pnum, seb->ec);
742 return seb;
743 }
744 }
745
746 ubi_err("no eraseblocks found");
747 return ERR_PTR(-ENOSPC);
748}
749
750/**
751 * process_eb - read UBI headers, check them and add corresponding data
752 * to the scanning information.
753 * @ubi: UBI device description object
754 * @si: scanning information
755 * @pnum: the physical eraseblock number
756 *
757 * This function returns a zero if the physical eraseblock was succesfully
758 * handled and a negative error code in case of failure.
759 */
760static int process_eb(struct ubi_device *ubi, struct ubi_scan_info *si, int pnum)
761{
762 long long ec;
763 int err, bitflips = 0, vol_id, ec_corr = 0;
764
765 dbg_bld("scan PEB %d", pnum);
766
767 /* Skip bad physical eraseblocks */
768 err = ubi_io_is_bad(ubi, pnum);
769 if (err < 0)
770 return err;
771 else if (err) {
772 /*
773 * FIXME: this is actually duty of the I/O unit to initialize
774 * this, but MTD does not provide enough information.
775 */
776 si->bad_peb_count += 1;
777 return 0;
778 }
779
780 err = ubi_io_read_ec_hdr(ubi, pnum, ech, 0);
781 if (err < 0)
782 return err;
783 else if (err == UBI_IO_BITFLIPS)
784 bitflips = 1;
785 else if (err == UBI_IO_PEB_EMPTY)
786 return ubi_scan_add_to_list(si, pnum, UBI_SCAN_UNKNOWN_EC,
787 &si->erase);
788 else if (err == UBI_IO_BAD_EC_HDR) {
789 /*
790 * We have to also look at the VID header, possibly it is not
791 * corrupted. Set %bitflips flag in order to make this PEB be
792 * moved and EC be re-created.
793 */
794 ec_corr = 1;
795 ec = UBI_SCAN_UNKNOWN_EC;
796 bitflips = 1;
797 }
798
799 si->is_empty = 0;
800
801 if (!ec_corr) {
802 /* Make sure UBI version is OK */
803 if (ech->version != UBI_VERSION) {
804 ubi_err("this UBI version is %d, image version is %d",
805 UBI_VERSION, (int)ech->version);
806 return -EINVAL;
807 }
808
809 ec = ubi64_to_cpu(ech->ec);
810 if (ec > UBI_MAX_ERASECOUNTER) {
811 /*
812 * Erase counter overflow. The EC headers have 64 bits
813 * reserved, but we anyway make use of only 31 bit
814 * values, as this seems to be enough for any existing
815 * flash. Upgrade UBI and use 64-bit erase counters
816 * internally.
817 */
818 ubi_err("erase counter overflow, max is %d",
819 UBI_MAX_ERASECOUNTER);
820 ubi_dbg_dump_ec_hdr(ech);
821 return -EINVAL;
822 }
823 }
824
825 /* OK, we've done with the EC header, let's look at the VID header */
826
827 err = ubi_io_read_vid_hdr(ubi, pnum, vidh, 0);
828 if (err < 0)
829 return err;
830 else if (err == UBI_IO_BITFLIPS)
831 bitflips = 1;
832 else if (err == UBI_IO_BAD_VID_HDR ||
833 (err == UBI_IO_PEB_FREE && ec_corr)) {
834 /* VID header is corrupted */
835 err = ubi_scan_add_to_list(si, pnum, ec, &si->corr);
836 if (err)
837 return err;
838 goto adjust_mean_ec;
839 } else if (err == UBI_IO_PEB_FREE) {
840 /* No VID header - the physical eraseblock is free */
841 err = ubi_scan_add_to_list(si, pnum, ec, &si->free);
842 if (err)
843 return err;
844 goto adjust_mean_ec;
845 }
846
847 vol_id = ubi32_to_cpu(vidh->vol_id);
848 if (vol_id > UBI_MAX_VOLUMES && vol_id != UBI_LAYOUT_VOL_ID) {
849 int lnum = ubi32_to_cpu(vidh->lnum);
850
851 /* Unsupported internal volume */
852 switch (vidh->compat) {
853 case UBI_COMPAT_DELETE:
854 ubi_msg("\"delete\" compatible internal volume %d:%d"
855 " found, remove it", vol_id, lnum);
856 err = ubi_scan_add_to_list(si, pnum, ec, &si->corr);
857 if (err)
858 return err;
859 break;
860
861 case UBI_COMPAT_RO:
862 ubi_msg("read-only compatible internal volume %d:%d"
863 " found, switch to read-only mode",
864 vol_id, lnum);
865 ubi->ro_mode = 1;
866 break;
867
868 case UBI_COMPAT_PRESERVE:
869 ubi_msg("\"preserve\" compatible internal volume %d:%d"
870 " found", vol_id, lnum);
871 err = ubi_scan_add_to_list(si, pnum, ec, &si->alien);
872 if (err)
873 return err;
874 si->alien_peb_count += 1;
875 return 0;
876
877 case UBI_COMPAT_REJECT:
878 ubi_err("incompatible internal volume %d:%d found",
879 vol_id, lnum);
880 return -EINVAL;
881 }
882 }
883
884 /* Both UBI headers seem to be fine */
885 err = ubi_scan_add_used(ubi, si, pnum, ec, vidh, bitflips);
886 if (err)
887 return err;
888
889adjust_mean_ec:
890 if (!ec_corr) {
891 if (si->ec_sum + ec < ec) {
892 commit_to_mean_value(si);
893 si->ec_sum = 0;
894 si->ec_count = 0;
895 } else {
896 si->ec_sum += ec;
897 si->ec_count += 1;
898 }
899
900 if (ec > si->max_ec)
901 si->max_ec = ec;
902 if (ec < si->min_ec)
903 si->min_ec = ec;
904 }
905
906 return 0;
907}
908
909/**
910 * ubi_scan - scan an MTD device.
911 * @ubi: UBI device description object
912 *
913 * This function does full scanning of an MTD device and returns complete
914 * information about it. In case of failure, an error code is returned.
915 */
916struct ubi_scan_info *ubi_scan(struct ubi_device *ubi)
917{
918 int err, pnum;
919 struct rb_node *rb1, *rb2;
920 struct ubi_scan_volume *sv;
921 struct ubi_scan_leb *seb;
922 struct ubi_scan_info *si;
923
924 si = kzalloc(sizeof(struct ubi_scan_info), GFP_KERNEL);
925 if (!si)
926 return ERR_PTR(-ENOMEM);
927
928 INIT_LIST_HEAD(&si->corr);
929 INIT_LIST_HEAD(&si->free);
930 INIT_LIST_HEAD(&si->erase);
931 INIT_LIST_HEAD(&si->alien);
932 si->volumes = RB_ROOT;
933 si->is_empty = 1;
934
935 err = -ENOMEM;
936 ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
937 if (!ech)
938 goto out_si;
939
940 vidh = ubi_zalloc_vid_hdr(ubi);
941 if (!vidh)
942 goto out_ech;
943
944 for (pnum = 0; pnum < ubi->peb_count; pnum++) {
945 cond_resched();
946
947 dbg_msg("process PEB %d", pnum);
948 err = process_eb(ubi, si, pnum);
949 if (err < 0)
950 goto out_vidh;
951 }
952
953 dbg_msg("scanning is finished");
954
955 /* Finish mean erase counter calculations */
956 if (si->ec_count)
957 commit_to_mean_value(si);
958
959 if (si->is_empty)
960 ubi_msg("empty MTD device detected");
961
962 /*
963 * In case of unknown erase counter we use the mean erase counter
964 * value.
965 */
966 ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
967 ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb)
968 if (seb->ec == UBI_SCAN_UNKNOWN_EC)
969 seb->ec = si->mean_ec;
970 }
971
972 list_for_each_entry(seb, &si->free, u.list) {
973 if (seb->ec == UBI_SCAN_UNKNOWN_EC)
974 seb->ec = si->mean_ec;
975 }
976
977 list_for_each_entry(seb, &si->corr, u.list)
978 if (seb->ec == UBI_SCAN_UNKNOWN_EC)
979 seb->ec = si->mean_ec;
980
981 list_for_each_entry(seb, &si->erase, u.list)
982 if (seb->ec == UBI_SCAN_UNKNOWN_EC)
983 seb->ec = si->mean_ec;
984
985 err = paranoid_check_si(ubi, si);
986 if (err) {
987 if (err > 0)
988 err = -EINVAL;
989 goto out_vidh;
990 }
991
992 ubi_free_vid_hdr(ubi, vidh);
993 kfree(ech);
994
995 return si;
996
997out_vidh:
998 ubi_free_vid_hdr(ubi, vidh);
999out_ech:
1000 kfree(ech);
1001out_si:
1002 ubi_scan_destroy_si(si);
1003 return ERR_PTR(err);
1004}
1005
1006/**
1007 * destroy_sv - free the scanning volume information
1008 * @sv: scanning volume information
1009 *
1010 * This function destroys the volume RB-tree (@sv->root) and the scanning
1011 * volume information.
1012 */
1013static void destroy_sv(struct ubi_scan_volume *sv)
1014{
1015 struct ubi_scan_leb *seb;
1016 struct rb_node *this = sv->root.rb_node;
1017
1018 while (this) {
1019 if (this->rb_left)
1020 this = this->rb_left;
1021 else if (this->rb_right)
1022 this = this->rb_right;
1023 else {
1024 seb = rb_entry(this, struct ubi_scan_leb, u.rb);
1025 this = rb_parent(this);
1026 if (this) {
1027 if (this->rb_left == &seb->u.rb)
1028 this->rb_left = NULL;
1029 else
1030 this->rb_right = NULL;
1031 }
1032
1033 kfree(seb);
1034 }
1035 }
1036 kfree(sv);
1037}
1038
1039/**
1040 * ubi_scan_destroy_si - destroy scanning information.
1041 * @si: scanning information
1042 */
1043void ubi_scan_destroy_si(struct ubi_scan_info *si)
1044{
1045 struct ubi_scan_leb *seb, *seb_tmp;
1046 struct ubi_scan_volume *sv;
1047 struct rb_node *rb;
1048
1049 list_for_each_entry_safe(seb, seb_tmp, &si->alien, u.list) {
1050 list_del(&seb->u.list);
1051 kfree(seb);
1052 }
1053 list_for_each_entry_safe(seb, seb_tmp, &si->erase, u.list) {
1054 list_del(&seb->u.list);
1055 kfree(seb);
1056 }
1057 list_for_each_entry_safe(seb, seb_tmp, &si->corr, u.list) {
1058 list_del(&seb->u.list);
1059 kfree(seb);
1060 }
1061 list_for_each_entry_safe(seb, seb_tmp, &si->free, u.list) {
1062 list_del(&seb->u.list);
1063 kfree(seb);
1064 }
1065
1066 /* Destroy the volume RB-tree */
1067 rb = si->volumes.rb_node;
1068 while (rb) {
1069 if (rb->rb_left)
1070 rb = rb->rb_left;
1071 else if (rb->rb_right)
1072 rb = rb->rb_right;
1073 else {
1074 sv = rb_entry(rb, struct ubi_scan_volume, rb);
1075
1076 rb = rb_parent(rb);
1077 if (rb) {
1078 if (rb->rb_left == &sv->rb)
1079 rb->rb_left = NULL;
1080 else
1081 rb->rb_right = NULL;
1082 }
1083
1084 destroy_sv(sv);
1085 }
1086 }
1087
1088 kfree(si);
1089}
1090
1091#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
1092
1093/**
1094 * paranoid_check_si - check if the scanning information is correct and
1095 * consistent.
1096 * @ubi: UBI device description object
1097 * @si: scanning information
1098 *
1099 * This function returns zero if the scanning information is all right, %1 if
1100 * not and a negative error code if an error occurred.
1101 */
1102static int paranoid_check_si(const struct ubi_device *ubi,
1103 struct ubi_scan_info *si)
1104{
1105 int pnum, err, vols_found = 0;
1106 struct rb_node *rb1, *rb2;
1107 struct ubi_scan_volume *sv;
1108 struct ubi_scan_leb *seb, *last_seb;
1109 uint8_t *buf;
1110
1111 /*
1112 * At first, check that scanning information is ok.
1113 */
1114 ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
1115 int leb_count = 0;
1116
1117 cond_resched();
1118
1119 vols_found += 1;
1120
1121 if (si->is_empty) {
1122 ubi_err("bad is_empty flag");
1123 goto bad_sv;
1124 }
1125
1126 if (sv->vol_id < 0 || sv->highest_lnum < 0 ||
1127 sv->leb_count < 0 || sv->vol_type < 0 || sv->used_ebs < 0 ||
1128 sv->data_pad < 0 || sv->last_data_size < 0) {
1129 ubi_err("negative values");
1130 goto bad_sv;
1131 }
1132
1133 if (sv->vol_id >= UBI_MAX_VOLUMES &&
1134 sv->vol_id < UBI_INTERNAL_VOL_START) {
1135 ubi_err("bad vol_id");
1136 goto bad_sv;
1137 }
1138
1139 if (sv->vol_id > si->highest_vol_id) {
1140 ubi_err("highest_vol_id is %d, but vol_id %d is there",
1141 si->highest_vol_id, sv->vol_id);
1142 goto out;
1143 }
1144
1145 if (sv->vol_type != UBI_DYNAMIC_VOLUME &&
1146 sv->vol_type != UBI_STATIC_VOLUME) {
1147 ubi_err("bad vol_type");
1148 goto bad_sv;
1149 }
1150
1151 if (sv->data_pad > ubi->leb_size / 2) {
1152 ubi_err("bad data_pad");
1153 goto bad_sv;
1154 }
1155
1156 last_seb = NULL;
1157 ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) {
1158 cond_resched();
1159
1160 last_seb = seb;
1161 leb_count += 1;
1162
1163 if (seb->pnum < 0 || seb->ec < 0) {
1164 ubi_err("negative values");
1165 goto bad_seb;
1166 }
1167
1168 if (seb->ec < si->min_ec) {
1169 ubi_err("bad si->min_ec (%d), %d found",
1170 si->min_ec, seb->ec);
1171 goto bad_seb;
1172 }
1173
1174 if (seb->ec > si->max_ec) {
1175 ubi_err("bad si->max_ec (%d), %d found",
1176 si->max_ec, seb->ec);
1177 goto bad_seb;
1178 }
1179
1180 if (seb->pnum >= ubi->peb_count) {
1181 ubi_err("too high PEB number %d, total PEBs %d",
1182 seb->pnum, ubi->peb_count);
1183 goto bad_seb;
1184 }
1185
1186 if (sv->vol_type == UBI_STATIC_VOLUME) {
1187 if (seb->lnum >= sv->used_ebs) {
1188 ubi_err("bad lnum or used_ebs");
1189 goto bad_seb;
1190 }
1191 } else {
1192 if (sv->used_ebs != 0) {
1193 ubi_err("non-zero used_ebs");
1194 goto bad_seb;
1195 }
1196 }
1197
1198 if (seb->lnum > sv->highest_lnum) {
1199 ubi_err("incorrect highest_lnum or lnum");
1200 goto bad_seb;
1201 }
1202 }
1203
1204 if (sv->leb_count != leb_count) {
1205 ubi_err("bad leb_count, %d objects in the tree",
1206 leb_count);
1207 goto bad_sv;
1208 }
1209
1210 if (!last_seb)
1211 continue;
1212
1213 seb = last_seb;
1214
1215 if (seb->lnum != sv->highest_lnum) {
1216 ubi_err("bad highest_lnum");
1217 goto bad_seb;
1218 }
1219 }
1220
1221 if (vols_found != si->vols_found) {
1222 ubi_err("bad si->vols_found %d, should be %d",
1223 si->vols_found, vols_found);
1224 goto out;
1225 }
1226
1227 /* Check that scanning information is correct */
1228 ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
1229 last_seb = NULL;
1230 ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) {
1231 int vol_type;
1232
1233 cond_resched();
1234
1235 last_seb = seb;
1236
1237 err = ubi_io_read_vid_hdr(ubi, seb->pnum, vidh, 1);
1238 if (err && err != UBI_IO_BITFLIPS) {
1239 ubi_err("VID header is not OK (%d)", err);
1240 if (err > 0)
1241 err = -EIO;
1242 return err;
1243 }
1244
1245 vol_type = vidh->vol_type == UBI_VID_DYNAMIC ?
1246 UBI_DYNAMIC_VOLUME : UBI_STATIC_VOLUME;
1247 if (sv->vol_type != vol_type) {
1248 ubi_err("bad vol_type");
1249 goto bad_vid_hdr;
1250 }
1251
1252 if (seb->sqnum != ubi64_to_cpu(vidh->sqnum)) {
1253 ubi_err("bad sqnum %llu", seb->sqnum);
1254 goto bad_vid_hdr;
1255 }
1256
1257 if (sv->vol_id != ubi32_to_cpu(vidh->vol_id)) {
1258 ubi_err("bad vol_id %d", sv->vol_id);
1259 goto bad_vid_hdr;
1260 }
1261
1262 if (sv->compat != vidh->compat) {
1263 ubi_err("bad compat %d", vidh->compat);
1264 goto bad_vid_hdr;
1265 }
1266
1267 if (seb->lnum != ubi32_to_cpu(vidh->lnum)) {
1268 ubi_err("bad lnum %d", seb->lnum);
1269 goto bad_vid_hdr;
1270 }
1271
1272 if (sv->used_ebs != ubi32_to_cpu(vidh->used_ebs)) {
1273 ubi_err("bad used_ebs %d", sv->used_ebs);
1274 goto bad_vid_hdr;
1275 }
1276
1277 if (sv->data_pad != ubi32_to_cpu(vidh->data_pad)) {
1278 ubi_err("bad data_pad %d", sv->data_pad);
1279 goto bad_vid_hdr;
1280 }
1281
1282 if (seb->leb_ver != ubi32_to_cpu(vidh->leb_ver)) {
1283 ubi_err("bad leb_ver %u", seb->leb_ver);
1284 goto bad_vid_hdr;
1285 }
1286 }
1287
1288 if (!last_seb)
1289 continue;
1290
1291 if (sv->highest_lnum != ubi32_to_cpu(vidh->lnum)) {
1292 ubi_err("bad highest_lnum %d", sv->highest_lnum);
1293 goto bad_vid_hdr;
1294 }
1295
1296 if (sv->last_data_size != ubi32_to_cpu(vidh->data_size)) {
1297 ubi_err("bad last_data_size %d", sv->last_data_size);
1298 goto bad_vid_hdr;
1299 }
1300 }
1301
1302 /*
1303 * Make sure that all the physical eraseblocks are in one of the lists
1304 * or trees.
1305 */
1306 buf = kmalloc(ubi->peb_count, GFP_KERNEL);
1307 if (!buf)
1308 return -ENOMEM;
1309
1310 memset(buf, 1, ubi->peb_count);
1311 for (pnum = 0; pnum < ubi->peb_count; pnum++) {
1312 err = ubi_io_is_bad(ubi, pnum);
1313 if (err < 0)
1314 return err;
1315 else if (err)
1316 buf[pnum] = 0;
1317 }
1318
1319 ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb)
1320 ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb)
1321 buf[seb->pnum] = 0;
1322
1323 list_for_each_entry(seb, &si->free, u.list)
1324 buf[seb->pnum] = 0;
1325
1326 list_for_each_entry(seb, &si->corr, u.list)
1327 buf[seb->pnum] = 0;
1328
1329 list_for_each_entry(seb, &si->erase, u.list)
1330 buf[seb->pnum] = 0;
1331
1332 list_for_each_entry(seb, &si->alien, u.list)
1333 buf[seb->pnum] = 0;
1334
1335 err = 0;
1336 for (pnum = 0; pnum < ubi->peb_count; pnum++)
1337 if (buf[pnum]) {
1338 ubi_err("PEB %d is not referred", pnum);
1339 err = 1;
1340 }
1341
1342 kfree(buf);
1343 if (err)
1344 goto out;
1345 return 0;
1346
1347bad_seb:
1348 ubi_err("bad scanning information about LEB %d", seb->lnum);
1349 ubi_dbg_dump_seb(seb, 0);
1350 ubi_dbg_dump_sv(sv);
1351 goto out;
1352
1353bad_sv:
1354 ubi_err("bad scanning information about volume %d", sv->vol_id);
1355 ubi_dbg_dump_sv(sv);
1356 goto out;
1357
1358bad_vid_hdr:
1359 ubi_err("bad scanning information about volume %d", sv->vol_id);
1360 ubi_dbg_dump_sv(sv);
1361 ubi_dbg_dump_vid_hdr(vidh);
1362
1363out:
1364 ubi_dbg_dump_stack();
1365 return 1;
1366}
1367
1368#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
diff --git a/drivers/mtd/ubi/scan.h b/drivers/mtd/ubi/scan.h
new file mode 100644
index 000000000000..3949f6192c76
--- /dev/null
+++ b/drivers/mtd/ubi/scan.h
@@ -0,0 +1,167 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21#ifndef __UBI_SCAN_H__
22#define __UBI_SCAN_H__
23
24/* The erase counter value for this physical eraseblock is unknown */
25#define UBI_SCAN_UNKNOWN_EC (-1)
26
27/**
28 * struct ubi_scan_leb - scanning information about a physical eraseblock.
29 * @ec: erase counter (%UBI_SCAN_UNKNOWN_EC if it is unknown)
30 * @pnum: physical eraseblock number
31 * @lnum: logical eraseblock number
32 * @scrub: if this physical eraseblock needs scrubbing
33 * @sqnum: sequence number
34 * @u: unions RB-tree or @list links
35 * @u.rb: link in the per-volume RB-tree of &struct ubi_scan_leb objects
36 * @u.list: link in one of the eraseblock lists
37 * @leb_ver: logical eraseblock version (obsolete)
38 *
39 * One object of this type is allocated for each physical eraseblock during
40 * scanning.
41 */
42struct ubi_scan_leb {
43 int ec;
44 int pnum;
45 int lnum;
46 int scrub;
47 unsigned long long sqnum;
48 union {
49 struct rb_node rb;
50 struct list_head list;
51 } u;
52 uint32_t leb_ver;
53};
54
55/**
56 * struct ubi_scan_volume - scanning information about a volume.
57 * @vol_id: volume ID
58 * @highest_lnum: highest logical eraseblock number in this volume
59 * @leb_count: number of logical eraseblocks in this volume
60 * @vol_type: volume type
61 * @used_ebs: number of used logical eraseblocks in this volume (only for
62 * static volumes)
63 * @last_data_size: amount of data in the last logical eraseblock of this
64 * volume (always equivalent to the usable logical eraseblock size in case of
65 * dynamic volumes)
66 * @data_pad: how many bytes at the end of logical eraseblocks of this volume
67 * are not used (due to volume alignment)
68 * @compat: compatibility flags of this volume
69 * @rb: link in the volume RB-tree
70 * @root: root of the RB-tree containing all the eraseblock belonging to this
71 * volume (&struct ubi_scan_leb objects)
72 *
73 * One object of this type is allocated for each volume during scanning.
74 */
75struct ubi_scan_volume {
76 int vol_id;
77 int highest_lnum;
78 int leb_count;
79 int vol_type;
80 int used_ebs;
81 int last_data_size;
82 int data_pad;
83 int compat;
84 struct rb_node rb;
85 struct rb_root root;
86};
87
88/**
89 * struct ubi_scan_info - UBI scanning information.
90 * @volumes: root of the volume RB-tree
91 * @corr: list of corrupted physical eraseblocks
92 * @free: list of free physical eraseblocks
93 * @erase: list of physical eraseblocks which have to be erased
94 * @alien: list of physical eraseblocks which should not be used by UBI (e.g.,
95 * @bad_peb_count: count of bad physical eraseblocks
96 * those belonging to "preserve"-compatible internal volumes)
97 * @vols_found: number of volumes found during scanning
98 * @highest_vol_id: highest volume ID
99 * @alien_peb_count: count of physical eraseblocks in the @alien list
100 * @is_empty: flag indicating whether the MTD device is empty or not
101 * @min_ec: lowest erase counter value
102 * @max_ec: highest erase counter value
103 * @max_sqnum: highest sequence number value
104 * @mean_ec: mean erase counter value
105 * @ec_sum: a temporary variable used when calculating @mean_ec
106 * @ec_count: a temporary variable used when calculating @mean_ec
107 *
108 * This data structure contains the result of scanning and may be used by other
109 * UBI units to build final UBI data structures, further error-recovery and so
110 * on.
111 */
112struct ubi_scan_info {
113 struct rb_root volumes;
114 struct list_head corr;
115 struct list_head free;
116 struct list_head erase;
117 struct list_head alien;
118 int bad_peb_count;
119 int vols_found;
120 int highest_vol_id;
121 int alien_peb_count;
122 int is_empty;
123 int min_ec;
124 int max_ec;
125 unsigned long long max_sqnum;
126 int mean_ec;
127 int ec_sum;
128 int ec_count;
129};
130
131struct ubi_device;
132struct ubi_vid_hdr;
133
134/*
135 * ubi_scan_move_to_list - move a physical eraseblock from the volume tree to a
136 * list.
137 *
138 * @sv: volume scanning information
139 * @seb: scanning eraseblock infprmation
140 * @list: the list to move to
141 */
142static inline void ubi_scan_move_to_list(struct ubi_scan_volume *sv,
143 struct ubi_scan_leb *seb,
144 struct list_head *list)
145{
146 rb_erase(&seb->u.rb, &sv->root);
147 list_add_tail(&seb->u.list, list);
148}
149
150int ubi_scan_add_to_list(struct ubi_scan_info *si, int pnum, int ec,
151 struct list_head *list);
152int ubi_scan_add_used(const struct ubi_device *ubi, struct ubi_scan_info *si,
153 int pnum, int ec, const struct ubi_vid_hdr *vid_hdr,
154 int bitflips);
155struct ubi_scan_volume *ubi_scan_find_sv(const struct ubi_scan_info *si,
156 int vol_id);
157struct ubi_scan_leb *ubi_scan_find_seb(const struct ubi_scan_volume *sv,
158 int lnum);
159void ubi_scan_rm_volume(struct ubi_scan_info *si, struct ubi_scan_volume *sv);
160struct ubi_scan_leb *ubi_scan_get_free_peb(const struct ubi_device *ubi,
161 struct ubi_scan_info *si);
162int ubi_scan_erase_peb(const struct ubi_device *ubi,
163 const struct ubi_scan_info *si, int pnum, int ec);
164struct ubi_scan_info *ubi_scan(struct ubi_device *ubi);
165void ubi_scan_destroy_si(struct ubi_scan_info *si);
166
167#endif /* !__UBI_SCAN_H__ */
diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
new file mode 100644
index 000000000000..feb647f108f0
--- /dev/null
+++ b/drivers/mtd/ubi/ubi.h
@@ -0,0 +1,535 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
3 * Copyright (c) Nokia Corporation, 2006, 2007
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 * the GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 *
19 * Author: Artem Bityutskiy (Битюцкий Артём)
20 */
21
22#ifndef __UBI_UBI_H__
23#define __UBI_UBI_H__
24
25#include <linux/init.h>
26#include <linux/types.h>
27#include <linux/list.h>
28#include <linux/rbtree.h>
29#include <linux/sched.h>
30#include <linux/wait.h>
31#include <linux/mutex.h>
32#include <linux/rwsem.h>
33#include <linux/spinlock.h>
34#include <linux/fs.h>
35#include <linux/cdev.h>
36#include <linux/device.h>
37#include <linux/string.h>
38#include <linux/mtd/mtd.h>
39
40#include <mtd/ubi-header.h>
41#include <linux/mtd/ubi.h>
42
43#include "scan.h"
44#include "debug.h"
45
46/* Maximum number of supported UBI devices */
47#define UBI_MAX_DEVICES 32
48
49/* UBI name used for character devices, sysfs, etc */
50#define UBI_NAME_STR "ubi"
51
52/* Normal UBI messages */
53#define ubi_msg(fmt, ...) printk(KERN_NOTICE "UBI: " fmt "\n", ##__VA_ARGS__)
54/* UBI warning messages */
55#define ubi_warn(fmt, ...) printk(KERN_WARNING "UBI warning: %s: " fmt "\n", \
56 __FUNCTION__, ##__VA_ARGS__)
57/* UBI error messages */
58#define ubi_err(fmt, ...) printk(KERN_ERR "UBI error: %s: " fmt "\n", \
59 __FUNCTION__, ##__VA_ARGS__)
60
61/* Lowest number PEBs reserved for bad PEB handling */
62#define MIN_RESEVED_PEBS 2
63
64/* Background thread name pattern */
65#define UBI_BGT_NAME_PATTERN "ubi_bgt%dd"
66
67/* This marker in the EBA table means that the LEB is um-mapped */
68#define UBI_LEB_UNMAPPED -1
69
70/*
71 * In case of errors, UBI tries to repeat the operation several times before
72 * returning error. The below constant defines how many times UBI re-tries.
73 */
74#define UBI_IO_RETRIES 3
75
76/*
77 * Error codes returned by the I/O unit.
78 *
79 * UBI_IO_PEB_EMPTY: the physical eraseblock is empty, i.e. it contains only
80 * 0xFF bytes
81 * UBI_IO_PEB_FREE: the physical eraseblock is free, i.e. it contains only a
82 * valid erase counter header, and the rest are %0xFF bytes
83 * UBI_IO_BAD_EC_HDR: the erase counter header is corrupted (bad magic or CRC)
84 * UBI_IO_BAD_VID_HDR: the volume identifier header is corrupted (bad magic or
85 * CRC)
86 * UBI_IO_BITFLIPS: bit-flips were detected and corrected
87 */
88enum {
89 UBI_IO_PEB_EMPTY = 1,
90 UBI_IO_PEB_FREE,
91 UBI_IO_BAD_EC_HDR,
92 UBI_IO_BAD_VID_HDR,
93 UBI_IO_BITFLIPS
94};
95
96extern int ubi_devices_cnt;
97extern struct ubi_device *ubi_devices[];
98
99struct ubi_volume_desc;
100
101/**
102 * struct ubi_volume - UBI volume description data structure.
103 * @dev: device object to make use of the the Linux device model
104 * @cdev: character device object to create character device
105 * @ubi: reference to the UBI device description object
106 * @vol_id: volume ID
107 * @readers: number of users holding this volume in read-only mode
108 * @writers: number of users holding this volume in read-write mode
109 * @exclusive: whether somebody holds this volume in exclusive mode
110 * @removed: if the volume was removed
111 * @checked: if this static volume was checked
112 *
113 * @reserved_pebs: how many physical eraseblocks are reserved for this volume
114 * @vol_type: volume type (%UBI_DYNAMIC_VOLUME or %UBI_STATIC_VOLUME)
115 * @usable_leb_size: logical eraseblock size without padding
116 * @used_ebs: how many logical eraseblocks in this volume contain data
117 * @last_eb_bytes: how many bytes are stored in the last logical eraseblock
118 * @used_bytes: how many bytes of data this volume contains
119 * @upd_marker: non-zero if the update marker is set for this volume
120 * @corrupted: non-zero if the volume is corrupted (static volumes only)
121 * @alignment: volume alignment
122 * @data_pad: how many bytes are not used at the end of physical eraseblocks to
123 * satisfy the requested alignment
124 * @name_len: volume name length
125 * @name: volume name
126 *
127 * @updating: whether the volume is being updated
128 * @upd_ebs: how many eraseblocks are expected to be updated
129 * @upd_bytes: how many bytes are expected to be received
130 * @upd_received: how many update bytes were already received
131 * @upd_buf: update buffer which is used to collect update data
132 *
133 * @eba_tbl: EBA table of this volume (LEB->PEB mapping)
134 *
135 * @gluebi_desc: gluebi UBI volume descriptor
136 * @gluebi_refcount: reference count of the gluebi MTD device
137 * @gluebi_mtd: MTD device description object of the gluebi MTD device
138 *
139 * The @corrupted field indicates that the volume's contents is corrupted.
140 * Since UBI protects only static volumes, this field is not relevant to
141 * dynamic volumes - it is user's responsibility to assure their data
142 * integrity.
143 *
144 * The @upd_marker flag indicates that this volume is either being updated at
145 * the moment or is damaged because of an unclean reboot.
146 */
147struct ubi_volume {
148 struct device dev;
149 struct cdev cdev;
150 struct ubi_device *ubi;
151 int vol_id;
152 int readers;
153 int writers;
154 int exclusive;
155 int removed;
156 int checked;
157
158 int reserved_pebs;
159 int vol_type;
160 int usable_leb_size;
161 int used_ebs;
162 int last_eb_bytes;
163 long long used_bytes;
164 int upd_marker;
165 int corrupted;
166 int alignment;
167 int data_pad;
168 int name_len;
169 char name[UBI_VOL_NAME_MAX+1];
170
171 int updating;
172 int upd_ebs;
173 long long upd_bytes;
174 long long upd_received;
175 void *upd_buf;
176
177 int *eba_tbl;
178
179#ifdef CONFIG_MTD_UBI_GLUEBI
180 /* Gluebi-related stuff may be compiled out */
181 struct ubi_volume_desc *gluebi_desc;
182 int gluebi_refcount;
183 struct mtd_info gluebi_mtd;
184#endif
185};
186
187/**
188 * struct ubi_volume_desc - descriptor of the UBI volume returned when it is
189 * opened.
190 * @vol: reference to the corresponding volume description object
191 * @mode: open mode (%UBI_READONLY, %UBI_READWRITE, or %UBI_EXCLUSIVE)
192 */
193struct ubi_volume_desc {
194 struct ubi_volume *vol;
195 int mode;
196};
197
198struct ubi_wl_entry;
199
200/**
201 * struct ubi_device - UBI device description structure
202 * @dev: class device object to use the the Linux device model
203 * @cdev: character device object to create character device
204 * @ubi_num: UBI device number
205 * @ubi_name: UBI device name
206 * @major: character device major number
207 * @vol_count: number of volumes in this UBI device
208 * @volumes: volumes of this UBI device
209 * @volumes_lock: protects @volumes, @rsvd_pebs, @avail_pebs, beb_rsvd_pebs,
210 * @beb_rsvd_level, @bad_peb_count, @good_peb_count, @vol_count, @vol->readers,
211 * @vol->writers, @vol->exclusive, @vol->removed, @vol->mapping and
212 * @vol->eba_tbl.
213 *
214 * @rsvd_pebs: count of reserved physical eraseblocks
215 * @avail_pebs: count of available physical eraseblocks
216 * @beb_rsvd_pebs: how many physical eraseblocks are reserved for bad PEB
217 * handling
218 * @beb_rsvd_level: normal level of PEBs reserved for bad PEB handling
219 *
220 * @vtbl_slots: how many slots are available in the volume table
221 * @vtbl_size: size of the volume table in bytes
222 * @vtbl: in-RAM volume table copy
223 *
224 * @max_ec: current highest erase counter value
225 * @mean_ec: current mean erase counter value
226 *
227 * global_sqnum: global sequence number
228 * @ltree_lock: protects the lock tree and @global_sqnum
229 * @ltree: the lock tree
230 * @vtbl_mutex: protects on-flash volume table
231 *
232 * @used: RB-tree of used physical eraseblocks
233 * @free: RB-tree of free physical eraseblocks
234 * @scrub: RB-tree of physical eraseblocks which need scrubbing
235 * @prot: protection trees
236 * @prot.pnum: protection tree indexed by physical eraseblock numbers
237 * @prot.aec: protection tree indexed by absolute erase counter value
238 * @wl_lock: protects the @used, @free, @prot, @lookuptbl, @abs_ec, @move_from,
239 * @move_to, @move_to_put @erase_pending, @wl_scheduled, and @works
240 * fields
241 * @wl_scheduled: non-zero if the wear-leveling was scheduled
242 * @lookuptbl: a table to quickly find a &struct ubi_wl_entry object for any
243 * physical eraseblock
244 * @abs_ec: absolute erase counter
245 * @move_from: physical eraseblock from where the data is being moved
246 * @move_to: physical eraseblock where the data is being moved to
247 * @move_from_put: if the "from" PEB was put
248 * @move_to_put: if the "to" PEB was put
249 * @works: list of pending works
250 * @works_count: count of pending works
251 * @bgt_thread: background thread description object
252 * @thread_enabled: if the background thread is enabled
253 * @bgt_name: background thread name
254 *
255 * @flash_size: underlying MTD device size (in bytes)
256 * @peb_count: count of physical eraseblocks on the MTD device
257 * @peb_size: physical eraseblock size
258 * @bad_peb_count: count of bad physical eraseblocks
259 * @good_peb_count: count of good physical eraseblocks
260 * @min_io_size: minimal input/output unit size of the underlying MTD device
261 * @hdrs_min_io_size: minimal I/O unit size used for VID and EC headers
262 * @ro_mode: if the UBI device is in read-only mode
263 * @leb_size: logical eraseblock size
264 * @leb_start: starting offset of logical eraseblocks within physical
265 * eraseblocks
266 * @ec_hdr_alsize: size of the EC header aligned to @hdrs_min_io_size
267 * @vid_hdr_alsize: size of the VID header aligned to @hdrs_min_io_size
268 * @vid_hdr_offset: starting offset of the volume identifier header (might be
269 * unaligned)
270 * @vid_hdr_aloffset: starting offset of the VID header aligned to
271 * @hdrs_min_io_size
272 * @vid_hdr_shift: contains @vid_hdr_offset - @vid_hdr_aloffset
273 * @bad_allowed: whether the MTD device admits of bad physical eraseblocks or
274 * not
275 * @mtd: MTD device descriptor
276 */
277struct ubi_device {
278 struct cdev cdev;
279 struct device dev;
280 int ubi_num;
281 char ubi_name[sizeof(UBI_NAME_STR)+5];
282 int major;
283 int vol_count;
284 struct ubi_volume *volumes[UBI_MAX_VOLUMES+UBI_INT_VOL_COUNT];
285 spinlock_t volumes_lock;
286
287 int rsvd_pebs;
288 int avail_pebs;
289 int beb_rsvd_pebs;
290 int beb_rsvd_level;
291
292 int vtbl_slots;
293 int vtbl_size;
294 struct ubi_vtbl_record *vtbl;
295 struct mutex vtbl_mutex;
296
297 int max_ec;
298 int mean_ec;
299
300 /* EBA unit's stuff */
301 unsigned long long global_sqnum;
302 spinlock_t ltree_lock;
303 struct rb_root ltree;
304
305 /* Wear-leveling unit's stuff */
306 struct rb_root used;
307 struct rb_root free;
308 struct rb_root scrub;
309 struct {
310 struct rb_root pnum;
311 struct rb_root aec;
312 } prot;
313 spinlock_t wl_lock;
314 int wl_scheduled;
315 struct ubi_wl_entry **lookuptbl;
316 unsigned long long abs_ec;
317 struct ubi_wl_entry *move_from;
318 struct ubi_wl_entry *move_to;
319 int move_from_put;
320 int move_to_put;
321 struct list_head works;
322 int works_count;
323 struct task_struct *bgt_thread;
324 int thread_enabled;
325 char bgt_name[sizeof(UBI_BGT_NAME_PATTERN)+2];
326
327 /* I/O unit's stuff */
328 long long flash_size;
329 int peb_count;
330 int peb_size;
331 int bad_peb_count;
332 int good_peb_count;
333 int min_io_size;
334 int hdrs_min_io_size;
335 int ro_mode;
336 int leb_size;
337 int leb_start;
338 int ec_hdr_alsize;
339 int vid_hdr_alsize;
340 int vid_hdr_offset;
341 int vid_hdr_aloffset;
342 int vid_hdr_shift;
343 int bad_allowed;
344 struct mtd_info *mtd;
345};
346
347extern struct file_operations ubi_cdev_operations;
348extern struct file_operations ubi_vol_cdev_operations;
349extern struct class *ubi_class;
350
351/* vtbl.c */
352int ubi_change_vtbl_record(struct ubi_device *ubi, int idx,
353 struct ubi_vtbl_record *vtbl_rec);
354int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si);
355
356/* vmt.c */
357int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req);
358int ubi_remove_volume(struct ubi_volume_desc *desc);
359int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs);
360int ubi_add_volume(struct ubi_device *ubi, int vol_id);
361void ubi_free_volume(struct ubi_device *ubi, int vol_id);
362
363/* upd.c */
364int ubi_start_update(struct ubi_device *ubi, int vol_id, long long bytes);
365int ubi_more_update_data(struct ubi_device *ubi, int vol_id,
366 const void __user *buf, int count);
367
368/* misc.c */
369int ubi_calc_data_len(const struct ubi_device *ubi, const void *buf, int length);
370int ubi_check_volume(struct ubi_device *ubi, int vol_id);
371void ubi_calculate_reserved(struct ubi_device *ubi);
372
373/* gluebi.c */
374#ifdef CONFIG_MTD_UBI_GLUEBI
375int ubi_create_gluebi(struct ubi_device *ubi, struct ubi_volume *vol);
376int ubi_destroy_gluebi(struct ubi_volume *vol);
377#else
378#define ubi_create_gluebi(ubi, vol) 0
379#define ubi_destroy_gluebi(vol) 0
380#endif
381
382/* eba.c */
383int ubi_eba_unmap_leb(struct ubi_device *ubi, int vol_id, int lnum);
384int ubi_eba_read_leb(struct ubi_device *ubi, int vol_id, int lnum, void *buf,
385 int offset, int len, int check);
386int ubi_eba_write_leb(struct ubi_device *ubi, int vol_id, int lnum,
387 const void *buf, int offset, int len, int dtype);
388int ubi_eba_write_leb_st(struct ubi_device *ubi, int vol_id, int lnum,
389 const void *buf, int len, int dtype,
390 int used_ebs);
391int ubi_eba_atomic_leb_change(struct ubi_device *ubi, int vol_id, int lnum,
392 const void *buf, int len, int dtype);
393int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
394 struct ubi_vid_hdr *vid_hdr);
395int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si);
396void ubi_eba_close(const struct ubi_device *ubi);
397
398/* wl.c */
399int ubi_wl_get_peb(struct ubi_device *ubi, int dtype);
400int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture);
401int ubi_wl_flush(struct ubi_device *ubi);
402int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum);
403int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si);
404void ubi_wl_close(struct ubi_device *ubi);
405
406/* io.c */
407int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
408 int len);
409int ubi_io_write(const struct ubi_device *ubi, const void *buf, int pnum,
410 int offset, int len);
411int ubi_io_sync_erase(const struct ubi_device *ubi, int pnum, int torture);
412int ubi_io_is_bad(const struct ubi_device *ubi, int pnum);
413int ubi_io_mark_bad(const struct ubi_device *ubi, int pnum);
414int ubi_io_read_ec_hdr(const struct ubi_device *ubi, int pnum,
415 struct ubi_ec_hdr *ec_hdr, int verbose);
416int ubi_io_write_ec_hdr(const struct ubi_device *ubi, int pnum,
417 struct ubi_ec_hdr *ec_hdr);
418int ubi_io_read_vid_hdr(const struct ubi_device *ubi, int pnum,
419 struct ubi_vid_hdr *vid_hdr, int verbose);
420int ubi_io_write_vid_hdr(const struct ubi_device *ubi, int pnum,
421 struct ubi_vid_hdr *vid_hdr);
422
423/*
424 * ubi_rb_for_each_entry - walk an RB-tree.
425 * @rb: a pointer to type 'struct rb_node' to to use as a loop counter
426 * @pos: a pointer to RB-tree entry type to use as a loop counter
427 * @root: RB-tree's root
428 * @member: the name of the 'struct rb_node' within the RB-tree entry
429 */
430#define ubi_rb_for_each_entry(rb, pos, root, member) \
431 for (rb = rb_first(root), \
432 pos = (rb ? container_of(rb, typeof(*pos), member) : NULL); \
433 rb; \
434 rb = rb_next(rb), pos = container_of(rb, typeof(*pos), member))
435
436/**
437 * ubi_zalloc_vid_hdr - allocate a volume identifier header object.
438 * @ubi: UBI device description object
439 *
440 * This function returns a pointer to the newly allocated and zero-filled
441 * volume identifier header object in case of success and %NULL in case of
442 * failure.
443 */
444static inline struct ubi_vid_hdr *ubi_zalloc_vid_hdr(const struct ubi_device *ubi)
445{
446 void *vid_hdr;
447
448 vid_hdr = kzalloc(ubi->vid_hdr_alsize, GFP_KERNEL);
449 if (!vid_hdr)
450 return NULL;
451
452 /*
453 * VID headers may be stored at un-aligned flash offsets, so we shift
454 * the pointer.
455 */
456 return vid_hdr + ubi->vid_hdr_shift;
457}
458
459/**
460 * ubi_free_vid_hdr - free a volume identifier header object.
461 * @ubi: UBI device description object
462 * @vid_hdr: the object to free
463 */
464static inline void ubi_free_vid_hdr(const struct ubi_device *ubi,
465 struct ubi_vid_hdr *vid_hdr)
466{
467 void *p = vid_hdr;
468
469 if (!p)
470 return;
471
472 kfree(p - ubi->vid_hdr_shift);
473}
474
475/*
476 * This function is equivalent to 'ubi_io_read()', but @offset is relative to
477 * the beginning of the logical eraseblock, not to the beginning of the
478 * physical eraseblock.
479 */
480static inline int ubi_io_read_data(const struct ubi_device *ubi, void *buf,
481 int pnum, int offset, int len)
482{
483 ubi_assert(offset >= 0);
484 return ubi_io_read(ubi, buf, pnum, offset + ubi->leb_start, len);
485}
486
487/*
488 * This function is equivalent to 'ubi_io_write()', but @offset is relative to
489 * the beginning of the logical eraseblock, not to the beginning of the
490 * physical eraseblock.
491 */
492static inline int ubi_io_write_data(const struct ubi_device *ubi, const void *buf,
493 int pnum, int offset, int len)
494{
495 ubi_assert(offset >= 0);
496 return ubi_io_write(ubi, buf, pnum, offset + ubi->leb_start, len);
497}
498
499/**
500 * ubi_ro_mode - switch to read-only mode.
501 * @ubi: UBI device description object
502 */
503static inline void ubi_ro_mode(struct ubi_device *ubi)
504{
505 ubi->ro_mode = 1;
506 ubi_warn("switch to read-only mode");
507}
508
509/**
510 * vol_id2idx - get table index by volume ID.
511 * @ubi: UBI device description object
512 * @vol_id: volume ID
513 */
514static inline int vol_id2idx(const struct ubi_device *ubi, int vol_id)
515{
516 if (vol_id >= UBI_INTERNAL_VOL_START)
517 return vol_id - UBI_INTERNAL_VOL_START + ubi->vtbl_slots;
518 else
519 return vol_id;
520}
521
522/**
523 * idx2vol_id - get volume ID by table index.
524 * @ubi: UBI device description object
525 * @idx: table index
526 */
527static inline int idx2vol_id(const struct ubi_device *ubi, int idx)
528{
529 if (idx >= ubi->vtbl_slots)
530 return idx - ubi->vtbl_slots + UBI_INTERNAL_VOL_START;
531 else
532 return idx;
533}
534
535#endif /* !__UBI_UBI_H__ */
diff --git a/drivers/mtd/ubi/upd.c b/drivers/mtd/ubi/upd.c
new file mode 100644
index 000000000000..8925b977e3dc
--- /dev/null
+++ b/drivers/mtd/ubi/upd.c
@@ -0,0 +1,348 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
3 * Copyright (c) Nokia Corporation, 2006
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 * the GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 *
19 * Author: Artem Bityutskiy (Битюцкий Артём)
20 *
21 * Jan 2007: Alexander Schmidt, hacked per-volume update.
22 */
23
24/*
25 * This file contains implementation of the volume update functionality.
26 *
27 * The update operation is based on the per-volume update marker which is
28 * stored in the volume table. The update marker is set before the update
29 * starts, and removed after the update has been finished. So if the update was
30 * interrupted by an unclean re-boot or due to some other reasons, the update
31 * marker stays on the flash media and UBI finds it when it attaches the MTD
32 * device next time. If the update marker is set for a volume, the volume is
33 * treated as damaged and most I/O operations are prohibited. Only a new update
34 * operation is allowed.
35 *
36 * Note, in general it is possible to implement the update operation as a
37 * transaction with a roll-back capability.
38 */
39
40#include <linux/err.h>
41#include <asm/uaccess.h>
42#include <asm/div64.h>
43#include "ubi.h"
44
45/**
46 * set_update_marker - set update marker.
47 * @ubi: UBI device description object
48 * @vol_id: volume ID
49 *
50 * This function sets the update marker flag for volume @vol_id. Returns zero
51 * in case of success and a negative error code in case of failure.
52 */
53static int set_update_marker(struct ubi_device *ubi, int vol_id)
54{
55 int err;
56 struct ubi_vtbl_record vtbl_rec;
57 struct ubi_volume *vol = ubi->volumes[vol_id];
58
59 dbg_msg("set update marker for volume %d", vol_id);
60
61 if (vol->upd_marker) {
62 ubi_assert(ubi->vtbl[vol_id].upd_marker);
63 dbg_msg("already set");
64 return 0;
65 }
66
67 memcpy(&vtbl_rec, &ubi->vtbl[vol_id], sizeof(struct ubi_vtbl_record));
68 vtbl_rec.upd_marker = 1;
69
70 err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
71 vol->upd_marker = 1;
72 return err;
73}
74
75/**
76 * clear_update_marker - clear update marker.
77 * @ubi: UBI device description object
78 * @vol_id: volume ID
79 * @bytes: new data size in bytes
80 *
81 * This function clears the update marker for volume @vol_id, sets new volume
82 * data size and clears the "corrupted" flag (static volumes only). Returns
83 * zero in case of success and a negative error code in case of failure.
84 */
85static int clear_update_marker(struct ubi_device *ubi, int vol_id, long long bytes)
86{
87 int err;
88 uint64_t tmp;
89 struct ubi_vtbl_record vtbl_rec;
90 struct ubi_volume *vol = ubi->volumes[vol_id];
91
92 dbg_msg("clear update marker for volume %d", vol_id);
93
94 memcpy(&vtbl_rec, &ubi->vtbl[vol_id], sizeof(struct ubi_vtbl_record));
95 ubi_assert(vol->upd_marker && vtbl_rec.upd_marker);
96 vtbl_rec.upd_marker = 0;
97
98 if (vol->vol_type == UBI_STATIC_VOLUME) {
99 vol->corrupted = 0;
100 vol->used_bytes = tmp = bytes;
101 vol->last_eb_bytes = do_div(tmp, vol->usable_leb_size);
102 vol->used_ebs = tmp;
103 if (vol->last_eb_bytes)
104 vol->used_ebs += 1;
105 else
106 vol->last_eb_bytes = vol->usable_leb_size;
107 }
108
109 err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
110 vol->upd_marker = 0;
111 return err;
112}
113
114/**
115 * ubi_start_update - start volume update.
116 * @ubi: UBI device description object
117 * @vol_id: volume ID
118 * @bytes: update bytes
119 *
120 * This function starts volume update operation. If @bytes is zero, the volume
121 * is just wiped out. Returns zero in case of success and a negative error code
122 * in case of failure.
123 */
124int ubi_start_update(struct ubi_device *ubi, int vol_id, long long bytes)
125{
126 int i, err;
127 uint64_t tmp;
128 struct ubi_volume *vol = ubi->volumes[vol_id];
129
130 dbg_msg("start update of volume %d, %llu bytes", vol_id, bytes);
131 vol->updating = 1;
132
133 err = set_update_marker(ubi, vol_id);
134 if (err)
135 return err;
136
137 /* Before updating - wipe out the volume */
138 for (i = 0; i < vol->reserved_pebs; i++) {
139 err = ubi_eba_unmap_leb(ubi, vol_id, i);
140 if (err)
141 return err;
142 }
143
144 if (bytes == 0) {
145 err = clear_update_marker(ubi, vol_id, 0);
146 if (err)
147 return err;
148 err = ubi_wl_flush(ubi);
149 if (!err)
150 vol->updating = 0;
151 }
152
153 vol->upd_buf = kmalloc(ubi->leb_size, GFP_KERNEL);
154 if (!vol->upd_buf)
155 return -ENOMEM;
156
157 tmp = bytes;
158 vol->upd_ebs = !!do_div(tmp, vol->usable_leb_size);
159 vol->upd_ebs += tmp;
160 vol->upd_bytes = bytes;
161 vol->upd_received = 0;
162 return 0;
163}
164
165/**
166 * write_leb - write update data.
167 * @ubi: UBI device description object
168 * @vol_id: volume ID
169 * @lnum: logical eraseblock number
170 * @buf: data to write
171 * @len: data size
172 * @used_ebs: how many logical eraseblocks will this volume contain (static
173 * volumes only)
174 *
175 * This function writes update data to corresponding logical eraseblock. In
176 * case of dynamic volume, this function checks if the data contains 0xFF bytes
177 * at the end. If yes, the 0xFF bytes are cut and not written. So if the whole
178 * buffer contains only 0xFF bytes, the LEB is left unmapped.
179 *
180 * The reason why we skip the trailing 0xFF bytes in case of dynamic volume is
181 * that we want to make sure that more data may be appended to the logical
182 * eraseblock in future. Indeed, writing 0xFF bytes may have side effects and
183 * this PEB won't be writable anymore. So if one writes the file-system image
184 * to the UBI volume where 0xFFs mean free space - UBI makes sure this free
185 * space is writable after the update.
186 *
187 * We do not do this for static volumes because they are read-only. But this
188 * also cannot be done because we have to store per-LEB CRC and the correct
189 * data length.
190 *
191 * This function returns zero in case of success and a negative error code in
192 * case of failure.
193 */
194static int write_leb(struct ubi_device *ubi, int vol_id, int lnum, void *buf,
195 int len, int used_ebs)
196{
197 int err, l;
198 struct ubi_volume *vol = ubi->volumes[vol_id];
199
200 if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
201 l = ALIGN(len, ubi->min_io_size);
202 memset(buf + len, 0xFF, l - len);
203
204 l = ubi_calc_data_len(ubi, buf, l);
205 if (l == 0) {
206 dbg_msg("all %d bytes contain 0xFF - skip", len);
207 return 0;
208 }
209 if (len != l)
210 dbg_msg("skip last %d bytes (0xFF)", len - l);
211
212 err = ubi_eba_write_leb(ubi, vol_id, lnum, buf, 0, l,
213 UBI_UNKNOWN);
214 } else {
215 /*
216 * When writing static volume, and this is the last logical
217 * eraseblock, the length (@len) does not have to be aligned to
218 * the minimal flash I/O unit. The 'ubi_eba_write_leb_st()'
219 * function accepts exact (unaligned) length and stores it in
220 * the VID header. And it takes care of proper alignment by
221 * padding the buffer. Here we just make sure the padding will
222 * contain zeros, not random trash.
223 */
224 memset(buf + len, 0, vol->usable_leb_size - len);
225 err = ubi_eba_write_leb_st(ubi, vol_id, lnum, buf, len,
226 UBI_UNKNOWN, used_ebs);
227 }
228
229 return err;
230}
231
232/**
233 * ubi_more_update_data - write more update data.
234 * @vol: volume description object
235 * @buf: write data (user-space memory buffer)
236 * @count: how much bytes to write
237 *
238 * This function writes more data to the volume which is being updated. It may
239 * be called arbitrary number of times until all of the update data arrive.
240 * This function returns %0 in case of success, number of bytes written during
241 * the last call if the whole volume update was successfully finished, and a
242 * negative error code in case of failure.
243 */
244int ubi_more_update_data(struct ubi_device *ubi, int vol_id,
245 const void __user *buf, int count)
246{
247 uint64_t tmp;
248 struct ubi_volume *vol = ubi->volumes[vol_id];
249 int lnum, offs, err = 0, len, to_write = count;
250
251 dbg_msg("write %d of %lld bytes, %lld already passed",
252 count, vol->upd_bytes, vol->upd_received);
253
254 if (ubi->ro_mode)
255 return -EROFS;
256
257 tmp = vol->upd_received;
258 offs = do_div(tmp, vol->usable_leb_size);
259 lnum = tmp;
260
261 if (vol->upd_received + count > vol->upd_bytes)
262 to_write = count = vol->upd_bytes - vol->upd_received;
263
264 /*
265 * When updating volumes, we accumulate whole logical eraseblock of
266 * data and write it at once.
267 */
268 if (offs != 0) {
269 /*
270 * This is a write to the middle of the logical eraseblock. We
271 * copy the data to our update buffer and wait for more data or
272 * flush it if the whole eraseblock is written or the update
273 * is finished.
274 */
275
276 len = vol->usable_leb_size - offs;
277 if (len > count)
278 len = count;
279
280 err = copy_from_user(vol->upd_buf + offs, buf, len);
281 if (err)
282 return -EFAULT;
283
284 if (offs + len == vol->usable_leb_size ||
285 vol->upd_received + len == vol->upd_bytes) {
286 int flush_len = offs + len;
287
288 /*
289 * OK, we gathered either the whole eraseblock or this
290 * is the last chunk, it's time to flush the buffer.
291 */
292 ubi_assert(flush_len <= vol->usable_leb_size);
293 err = write_leb(ubi, vol_id, lnum, vol->upd_buf,
294 flush_len, vol->upd_ebs);
295 if (err)
296 return err;
297 }
298
299 vol->upd_received += len;
300 count -= len;
301 buf += len;
302 lnum += 1;
303 }
304
305 /*
306 * If we've got more to write, let's continue. At this point we know we
307 * are starting from the beginning of an eraseblock.
308 */
309 while (count) {
310 if (count > vol->usable_leb_size)
311 len = vol->usable_leb_size;
312 else
313 len = count;
314
315 err = copy_from_user(vol->upd_buf, buf, len);
316 if (err)
317 return -EFAULT;
318
319 if (len == vol->usable_leb_size ||
320 vol->upd_received + len == vol->upd_bytes) {
321 err = write_leb(ubi, vol_id, lnum, vol->upd_buf, len,
322 vol->upd_ebs);
323 if (err)
324 break;
325 }
326
327 vol->upd_received += len;
328 count -= len;
329 lnum += 1;
330 buf += len;
331 }
332
333 ubi_assert(vol->upd_received <= vol->upd_bytes);
334 if (vol->upd_received == vol->upd_bytes) {
335 /* The update is finished, clear the update marker */
336 err = clear_update_marker(ubi, vol_id, vol->upd_bytes);
337 if (err)
338 return err;
339 err = ubi_wl_flush(ubi);
340 if (err == 0) {
341 err = to_write;
342 kfree(vol->upd_buf);
343 vol->updating = 0;
344 }
345 }
346
347 return err;
348}
diff --git a/drivers/mtd/ubi/vmt.c b/drivers/mtd/ubi/vmt.c
new file mode 100644
index 000000000000..622d0d18952c
--- /dev/null
+++ b/drivers/mtd/ubi/vmt.c
@@ -0,0 +1,809 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/*
22 * This file contains implementation of volume creation, deletion, updating and
23 * resizing.
24 */
25
26#include <linux/err.h>
27#include <asm/div64.h>
28#include "ubi.h"
29
30#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
31static void paranoid_check_volumes(struct ubi_device *ubi);
32#else
33#define paranoid_check_volumes(ubi)
34#endif
35
36static ssize_t vol_attribute_show(struct device *dev,
37 struct device_attribute *attr, char *buf);
38
39/* Device attributes corresponding to files in '/<sysfs>/class/ubi/ubiX_Y' */
40static struct device_attribute vol_reserved_ebs =
41 __ATTR(reserved_ebs, S_IRUGO, vol_attribute_show, NULL);
42static struct device_attribute vol_type =
43 __ATTR(type, S_IRUGO, vol_attribute_show, NULL);
44static struct device_attribute vol_name =
45 __ATTR(name, S_IRUGO, vol_attribute_show, NULL);
46static struct device_attribute vol_corrupted =
47 __ATTR(corrupted, S_IRUGO, vol_attribute_show, NULL);
48static struct device_attribute vol_alignment =
49 __ATTR(alignment, S_IRUGO, vol_attribute_show, NULL);
50static struct device_attribute vol_usable_eb_size =
51 __ATTR(usable_eb_size, S_IRUGO, vol_attribute_show, NULL);
52static struct device_attribute vol_data_bytes =
53 __ATTR(data_bytes, S_IRUGO, vol_attribute_show, NULL);
54static struct device_attribute vol_upd_marker =
55 __ATTR(upd_marker, S_IRUGO, vol_attribute_show, NULL);
56
57/*
58 * "Show" method for files in '/<sysfs>/class/ubi/ubiX_Y/'.
59 *
60 * Consider a situation:
61 * A. process 1 opens a sysfs file related to volume Y, say
62 * /<sysfs>/class/ubi/ubiX_Y/reserved_ebs;
63 * B. process 2 removes volume Y;
64 * C. process 1 starts reading the /<sysfs>/class/ubi/ubiX_Y/reserved_ebs file;
65 *
66 * What we want to do in a situation like that is to return error when the file
67 * is read. This is done by means of the 'removed' flag and the 'vol_lock' of
68 * the UBI volume description object.
69 */
70static ssize_t vol_attribute_show(struct device *dev,
71 struct device_attribute *attr, char *buf)
72{
73 int ret;
74 struct ubi_volume *vol = container_of(dev, struct ubi_volume, dev);
75
76 spin_lock(&vol->ubi->volumes_lock);
77 if (vol->removed) {
78 spin_unlock(&vol->ubi->volumes_lock);
79 return -ENODEV;
80 }
81 if (attr == &vol_reserved_ebs)
82 ret = sprintf(buf, "%d\n", vol->reserved_pebs);
83 else if (attr == &vol_type) {
84 const char *tp;
85 tp = vol->vol_type == UBI_DYNAMIC_VOLUME ? "dynamic" : "static";
86 ret = sprintf(buf, "%s\n", tp);
87 } else if (attr == &vol_name)
88 ret = sprintf(buf, "%s\n", vol->name);
89 else if (attr == &vol_corrupted)
90 ret = sprintf(buf, "%d\n", vol->corrupted);
91 else if (attr == &vol_alignment)
92 ret = sprintf(buf, "%d\n", vol->alignment);
93 else if (attr == &vol_usable_eb_size) {
94 ret = sprintf(buf, "%d\n", vol->usable_leb_size);
95 } else if (attr == &vol_data_bytes)
96 ret = sprintf(buf, "%lld\n", vol->used_bytes);
97 else if (attr == &vol_upd_marker)
98 ret = sprintf(buf, "%d\n", vol->upd_marker);
99 else
100 BUG();
101 spin_unlock(&vol->ubi->volumes_lock);
102 return ret;
103}
104
105/* Release method for volume devices */
106static void vol_release(struct device *dev)
107{
108 struct ubi_volume *vol = container_of(dev, struct ubi_volume, dev);
109 ubi_assert(vol->removed);
110 kfree(vol);
111}
112
113/**
114 * volume_sysfs_init - initialize sysfs for new volume.
115 * @ubi: UBI device description object
116 * @vol: volume description object
117 *
118 * This function returns zero in case of success and a negative error code in
119 * case of failure.
120 *
121 * Note, this function does not free allocated resources in case of failure -
122 * the caller does it. This is because this would cause release() here and the
123 * caller would oops.
124 */
125static int volume_sysfs_init(struct ubi_device *ubi, struct ubi_volume *vol)
126{
127 int err;
128
129 err = device_create_file(&vol->dev, &vol_reserved_ebs);
130 if (err)
131 return err;
132 err = device_create_file(&vol->dev, &vol_type);
133 if (err)
134 return err;
135 err = device_create_file(&vol->dev, &vol_name);
136 if (err)
137 return err;
138 err = device_create_file(&vol->dev, &vol_corrupted);
139 if (err)
140 return err;
141 err = device_create_file(&vol->dev, &vol_alignment);
142 if (err)
143 return err;
144 err = device_create_file(&vol->dev, &vol_usable_eb_size);
145 if (err)
146 return err;
147 err = device_create_file(&vol->dev, &vol_data_bytes);
148 if (err)
149 return err;
150 err = device_create_file(&vol->dev, &vol_upd_marker);
151 if (err)
152 return err;
153 return 0;
154}
155
156/**
157 * volume_sysfs_close - close sysfs for a volume.
158 * @vol: volume description object
159 */
160static void volume_sysfs_close(struct ubi_volume *vol)
161{
162 device_remove_file(&vol->dev, &vol_upd_marker);
163 device_remove_file(&vol->dev, &vol_data_bytes);
164 device_remove_file(&vol->dev, &vol_usable_eb_size);
165 device_remove_file(&vol->dev, &vol_alignment);
166 device_remove_file(&vol->dev, &vol_corrupted);
167 device_remove_file(&vol->dev, &vol_name);
168 device_remove_file(&vol->dev, &vol_type);
169 device_remove_file(&vol->dev, &vol_reserved_ebs);
170 device_unregister(&vol->dev);
171}
172
173/**
174 * ubi_create_volume - create volume.
175 * @ubi: UBI device description object
176 * @req: volume creation request
177 *
178 * This function creates volume described by @req. If @req->vol_id id
179 * %UBI_VOL_NUM_AUTO, this function automatically assigne ID to the new volume
180 * and saves it in @req->vol_id. Returns zero in case of success and a negative
181 * error code in case of failure.
182 */
183int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
184{
185 int i, err, vol_id = req->vol_id;
186 struct ubi_volume *vol;
187 struct ubi_vtbl_record vtbl_rec;
188 uint64_t bytes;
189
190 if (ubi->ro_mode)
191 return -EROFS;
192
193 vol = kzalloc(sizeof(struct ubi_volume), GFP_KERNEL);
194 if (!vol)
195 return -ENOMEM;
196
197 spin_lock(&ubi->volumes_lock);
198
199 if (vol_id == UBI_VOL_NUM_AUTO) {
200 /* Find unused volume ID */
201 dbg_msg("search for vacant volume ID");
202 for (i = 0; i < ubi->vtbl_slots; i++)
203 if (!ubi->volumes[i]) {
204 vol_id = i;
205 break;
206 }
207
208 if (vol_id == UBI_VOL_NUM_AUTO) {
209 dbg_err("out of volume IDs");
210 err = -ENFILE;
211 goto out_unlock;
212 }
213 req->vol_id = vol_id;
214 }
215
216 dbg_msg("volume ID %d, %llu bytes, type %d, name %s",
217 vol_id, (unsigned long long)req->bytes,
218 (int)req->vol_type, req->name);
219
220 /* Ensure that this volume does not exist */
221 err = -EEXIST;
222 if (ubi->volumes[vol_id]) {
223 dbg_err("volume %d already exists", vol_id);
224 goto out_unlock;
225 }
226
227 /* Ensure that the name is unique */
228 for (i = 0; i < ubi->vtbl_slots; i++)
229 if (ubi->volumes[i] &&
230 ubi->volumes[i]->name_len == req->name_len &&
231 strcmp(ubi->volumes[i]->name, req->name) == 0) {
232 dbg_err("volume \"%s\" exists (ID %d)", req->name, i);
233 goto out_unlock;
234 }
235
236 /* Calculate how many eraseblocks are requested */
237 vol->usable_leb_size = ubi->leb_size - ubi->leb_size % req->alignment;
238 bytes = req->bytes;
239 if (do_div(bytes, vol->usable_leb_size))
240 vol->reserved_pebs = 1;
241 vol->reserved_pebs += bytes;
242
243 /* Reserve physical eraseblocks */
244 if (vol->reserved_pebs > ubi->avail_pebs) {
245 dbg_err("not enough PEBs, only %d available", ubi->avail_pebs);
246 spin_unlock(&ubi->volumes_lock);
247 err = -ENOSPC;
248 goto out_unlock;
249 }
250 ubi->avail_pebs -= vol->reserved_pebs;
251 ubi->rsvd_pebs += vol->reserved_pebs;
252
253 vol->vol_id = vol_id;
254 vol->alignment = req->alignment;
255 vol->data_pad = ubi->leb_size % vol->alignment;
256 vol->vol_type = req->vol_type;
257 vol->name_len = req->name_len;
258 memcpy(vol->name, req->name, vol->name_len + 1);
259 vol->exclusive = 1;
260 vol->ubi = ubi;
261 ubi->volumes[vol_id] = vol;
262 spin_unlock(&ubi->volumes_lock);
263
264 /*
265 * Finish all pending erases because there may be some LEBs belonging
266 * to the same volume ID.
267 */
268 err = ubi_wl_flush(ubi);
269 if (err)
270 goto out_acc;
271
272 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int), GFP_KERNEL);
273 if (!vol->eba_tbl) {
274 err = -ENOMEM;
275 goto out_acc;
276 }
277
278 for (i = 0; i < vol->reserved_pebs; i++)
279 vol->eba_tbl[i] = UBI_LEB_UNMAPPED;
280
281 if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
282 vol->used_ebs = vol->reserved_pebs;
283 vol->last_eb_bytes = vol->usable_leb_size;
284 vol->used_bytes = vol->used_ebs * vol->usable_leb_size;
285 } else {
286 bytes = vol->used_bytes;
287 vol->last_eb_bytes = do_div(bytes, vol->usable_leb_size);
288 vol->used_ebs = bytes;
289 if (vol->last_eb_bytes)
290 vol->used_ebs += 1;
291 else
292 vol->last_eb_bytes = vol->usable_leb_size;
293 }
294
295 /* Register character device for the volume */
296 cdev_init(&vol->cdev, &ubi_vol_cdev_operations);
297 vol->cdev.owner = THIS_MODULE;
298 err = cdev_add(&vol->cdev, MKDEV(ubi->major, vol_id + 1), 1);
299 if (err) {
300 ubi_err("cannot add character device for volume %d", vol_id);
301 goto out_mapping;
302 }
303
304 err = ubi_create_gluebi(ubi, vol);
305 if (err)
306 goto out_cdev;
307
308 vol->dev.release = vol_release;
309 vol->dev.parent = &ubi->dev;
310 vol->dev.devt = MKDEV(ubi->major, vol->vol_id + 1);
311 vol->dev.class = ubi_class;
312 sprintf(&vol->dev.bus_id[0], "%s_%d", ubi->ubi_name, vol->vol_id);
313 err = device_register(&vol->dev);
314 if (err)
315 goto out_gluebi;
316
317 err = volume_sysfs_init(ubi, vol);
318 if (err)
319 goto out_sysfs;
320
321 /* Fill volume table record */
322 memset(&vtbl_rec, 0, sizeof(struct ubi_vtbl_record));
323 vtbl_rec.reserved_pebs = cpu_to_ubi32(vol->reserved_pebs);
324 vtbl_rec.alignment = cpu_to_ubi32(vol->alignment);
325 vtbl_rec.data_pad = cpu_to_ubi32(vol->data_pad);
326 vtbl_rec.name_len = cpu_to_ubi16(vol->name_len);
327 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
328 vtbl_rec.vol_type = UBI_VID_DYNAMIC;
329 else
330 vtbl_rec.vol_type = UBI_VID_STATIC;
331 memcpy(vtbl_rec.name, vol->name, vol->name_len + 1);
332
333 err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
334 if (err)
335 goto out_sysfs;
336
337 spin_lock(&ubi->volumes_lock);
338 ubi->vol_count += 1;
339 vol->exclusive = 0;
340 spin_unlock(&ubi->volumes_lock);
341
342 paranoid_check_volumes(ubi);
343 return 0;
344
345out_gluebi:
346 err = ubi_destroy_gluebi(vol);
347out_cdev:
348 cdev_del(&vol->cdev);
349out_mapping:
350 kfree(vol->eba_tbl);
351out_acc:
352 spin_lock(&ubi->volumes_lock);
353 ubi->rsvd_pebs -= vol->reserved_pebs;
354 ubi->avail_pebs += vol->reserved_pebs;
355out_unlock:
356 spin_unlock(&ubi->volumes_lock);
357 kfree(vol);
358 return err;
359
360 /*
361 * We are registered, so @vol is destroyed in the release function and
362 * we have to de-initialize differently.
363 */
364out_sysfs:
365 err = ubi_destroy_gluebi(vol);
366 cdev_del(&vol->cdev);
367 kfree(vol->eba_tbl);
368 spin_lock(&ubi->volumes_lock);
369 ubi->rsvd_pebs -= vol->reserved_pebs;
370 ubi->avail_pebs += vol->reserved_pebs;
371 spin_unlock(&ubi->volumes_lock);
372 volume_sysfs_close(vol);
373 return err;
374}
375
376/**
377 * ubi_remove_volume - remove volume.
378 * @desc: volume descriptor
379 *
380 * This function removes volume described by @desc. The volume has to be opened
381 * in "exclusive" mode. Returns zero in case of success and a negative error
382 * code in case of failure.
383 */
384int ubi_remove_volume(struct ubi_volume_desc *desc)
385{
386 struct ubi_volume *vol = desc->vol;
387 struct ubi_device *ubi = vol->ubi;
388 int i, err, vol_id = vol->vol_id, reserved_pebs = vol->reserved_pebs;
389
390 dbg_msg("remove UBI volume %d", vol_id);
391 ubi_assert(desc->mode == UBI_EXCLUSIVE);
392 ubi_assert(vol == ubi->volumes[vol_id]);
393
394 if (ubi->ro_mode)
395 return -EROFS;
396
397 err = ubi_destroy_gluebi(vol);
398 if (err)
399 return err;
400
401 err = ubi_change_vtbl_record(ubi, vol_id, NULL);
402 if (err)
403 return err;
404
405 for (i = 0; i < vol->reserved_pebs; i++) {
406 err = ubi_eba_unmap_leb(ubi, vol_id, i);
407 if (err)
408 return err;
409 }
410
411 spin_lock(&ubi->volumes_lock);
412 vol->removed = 1;
413 ubi->volumes[vol_id] = NULL;
414 spin_unlock(&ubi->volumes_lock);
415
416 kfree(vol->eba_tbl);
417 vol->eba_tbl = NULL;
418 cdev_del(&vol->cdev);
419 volume_sysfs_close(vol);
420 kfree(desc);
421
422 spin_lock(&ubi->volumes_lock);
423 ubi->rsvd_pebs -= reserved_pebs;
424 ubi->avail_pebs += reserved_pebs;
425 i = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs;
426 if (i > 0) {
427 i = ubi->avail_pebs >= i ? i : ubi->avail_pebs;
428 ubi->avail_pebs -= i;
429 ubi->rsvd_pebs += i;
430 ubi->beb_rsvd_pebs += i;
431 if (i > 0)
432 ubi_msg("reserve more %d PEBs", i);
433 }
434 ubi->vol_count -= 1;
435 spin_unlock(&ubi->volumes_lock);
436
437 paranoid_check_volumes(ubi);
438 module_put(THIS_MODULE);
439 return 0;
440}
441
442/**
443 * ubi_resize_volume - re-size volume.
444 * @desc: volume descriptor
445 * @reserved_pebs: new size in physical eraseblocks
446 *
447 * This function returns zero in case of success, and a negative error code in
448 * case of failure.
449 */
450int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
451{
452 int i, err, pebs, *new_mapping;
453 struct ubi_volume *vol = desc->vol;
454 struct ubi_device *ubi = vol->ubi;
455 struct ubi_vtbl_record vtbl_rec;
456 int vol_id = vol->vol_id;
457
458 if (ubi->ro_mode)
459 return -EROFS;
460
461 dbg_msg("re-size volume %d to from %d to %d PEBs",
462 vol_id, vol->reserved_pebs, reserved_pebs);
463 ubi_assert(desc->mode == UBI_EXCLUSIVE);
464 ubi_assert(vol == ubi->volumes[vol_id]);
465
466 if (vol->vol_type == UBI_STATIC_VOLUME &&
467 reserved_pebs < vol->used_ebs) {
468 dbg_err("too small size %d, %d LEBs contain data",
469 reserved_pebs, vol->used_ebs);
470 return -EINVAL;
471 }
472
473 /* If the size is the same, we have nothing to do */
474 if (reserved_pebs == vol->reserved_pebs)
475 return 0;
476
477 new_mapping = kmalloc(reserved_pebs * sizeof(int), GFP_KERNEL);
478 if (!new_mapping)
479 return -ENOMEM;
480
481 for (i = 0; i < reserved_pebs; i++)
482 new_mapping[i] = UBI_LEB_UNMAPPED;
483
484 /* Reserve physical eraseblocks */
485 pebs = reserved_pebs - vol->reserved_pebs;
486 if (pebs > 0) {
487 spin_lock(&ubi->volumes_lock);
488 if (pebs > ubi->avail_pebs) {
489 dbg_err("not enough PEBs: requested %d, available %d",
490 pebs, ubi->avail_pebs);
491 spin_unlock(&ubi->volumes_lock);
492 err = -ENOSPC;
493 goto out_free;
494 }
495 ubi->avail_pebs -= pebs;
496 ubi->rsvd_pebs += pebs;
497 for (i = 0; i < vol->reserved_pebs; i++)
498 new_mapping[i] = vol->eba_tbl[i];
499 kfree(vol->eba_tbl);
500 vol->eba_tbl = new_mapping;
501 spin_unlock(&ubi->volumes_lock);
502 }
503
504 /* Change volume table record */
505 memcpy(&vtbl_rec, &ubi->vtbl[vol_id], sizeof(struct ubi_vtbl_record));
506 vtbl_rec.reserved_pebs = cpu_to_ubi32(reserved_pebs);
507 err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
508 if (err)
509 goto out_acc;
510
511 if (pebs < 0) {
512 for (i = 0; i < -pebs; i++) {
513 err = ubi_eba_unmap_leb(ubi, vol_id, reserved_pebs + i);
514 if (err)
515 goto out_acc;
516 }
517 spin_lock(&ubi->volumes_lock);
518 ubi->rsvd_pebs += pebs;
519 ubi->avail_pebs -= pebs;
520 pebs = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs;
521 if (pebs > 0) {
522 pebs = ubi->avail_pebs >= pebs ? pebs : ubi->avail_pebs;
523 ubi->avail_pebs -= pebs;
524 ubi->rsvd_pebs += pebs;
525 ubi->beb_rsvd_pebs += pebs;
526 if (pebs > 0)
527 ubi_msg("reserve more %d PEBs", pebs);
528 }
529 for (i = 0; i < reserved_pebs; i++)
530 new_mapping[i] = vol->eba_tbl[i];
531 kfree(vol->eba_tbl);
532 vol->eba_tbl = new_mapping;
533 spin_unlock(&ubi->volumes_lock);
534 }
535
536 vol->reserved_pebs = reserved_pebs;
537 if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
538 vol->used_ebs = reserved_pebs;
539 vol->last_eb_bytes = vol->usable_leb_size;
540 vol->used_bytes = vol->used_ebs * vol->usable_leb_size;
541 }
542
543 paranoid_check_volumes(ubi);
544 return 0;
545
546out_acc:
547 if (pebs > 0) {
548 spin_lock(&ubi->volumes_lock);
549 ubi->rsvd_pebs -= pebs;
550 ubi->avail_pebs += pebs;
551 spin_unlock(&ubi->volumes_lock);
552 }
553out_free:
554 kfree(new_mapping);
555 return err;
556}
557
558/**
559 * ubi_add_volume - add volume.
560 * @ubi: UBI device description object
561 * @vol_id: volume ID
562 *
563 * This function adds an existin volume and initializes all its data
564 * structures. Returnes zero in case of success and a negative error code in
565 * case of failure.
566 */
567int ubi_add_volume(struct ubi_device *ubi, int vol_id)
568{
569 int err;
570 struct ubi_volume *vol = ubi->volumes[vol_id];
571
572 dbg_msg("add volume %d", vol_id);
573 ubi_dbg_dump_vol_info(vol);
574 ubi_assert(vol);
575
576 /* Register character device for the volume */
577 cdev_init(&vol->cdev, &ubi_vol_cdev_operations);
578 vol->cdev.owner = THIS_MODULE;
579 err = cdev_add(&vol->cdev, MKDEV(ubi->major, vol->vol_id + 1), 1);
580 if (err) {
581 ubi_err("cannot add character device for volume %d", vol_id);
582 return err;
583 }
584
585 err = ubi_create_gluebi(ubi, vol);
586 if (err)
587 goto out_cdev;
588
589 vol->dev.release = vol_release;
590 vol->dev.parent = &ubi->dev;
591 vol->dev.devt = MKDEV(ubi->major, vol->vol_id + 1);
592 vol->dev.class = ubi_class;
593 sprintf(&vol->dev.bus_id[0], "%s_%d", ubi->ubi_name, vol->vol_id);
594 err = device_register(&vol->dev);
595 if (err)
596 goto out_gluebi;
597
598 err = volume_sysfs_init(ubi, vol);
599 if (err) {
600 cdev_del(&vol->cdev);
601 err = ubi_destroy_gluebi(vol);
602 volume_sysfs_close(vol);
603 return err;
604 }
605
606 paranoid_check_volumes(ubi);
607 return 0;
608
609out_gluebi:
610 err = ubi_destroy_gluebi(vol);
611out_cdev:
612 cdev_del(&vol->cdev);
613 return err;
614}
615
616/**
617 * ubi_free_volume - free volume.
618 * @ubi: UBI device description object
619 * @vol_id: volume ID
620 *
621 * This function frees all resources for volume @vol_id but does not remove it.
622 * Used only when the UBI device is detached.
623 */
624void ubi_free_volume(struct ubi_device *ubi, int vol_id)
625{
626 int err;
627 struct ubi_volume *vol = ubi->volumes[vol_id];
628
629 dbg_msg("free volume %d", vol_id);
630 ubi_assert(vol);
631
632 vol->removed = 1;
633 err = ubi_destroy_gluebi(vol);
634 ubi->volumes[vol_id] = NULL;
635 cdev_del(&vol->cdev);
636 volume_sysfs_close(vol);
637}
638
639#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
640
641/**
642 * paranoid_check_volume - check volume information.
643 * @ubi: UBI device description object
644 * @vol_id: volume ID
645 */
646static void paranoid_check_volume(const struct ubi_device *ubi, int vol_id)
647{
648 int idx = vol_id2idx(ubi, vol_id);
649 int reserved_pebs, alignment, data_pad, vol_type, name_len, upd_marker;
650 const struct ubi_volume *vol = ubi->volumes[idx];
651 long long n;
652 const char *name;
653
654 reserved_pebs = ubi32_to_cpu(ubi->vtbl[vol_id].reserved_pebs);
655
656 if (!vol) {
657 if (reserved_pebs) {
658 ubi_err("no volume info, but volume exists");
659 goto fail;
660 }
661 return;
662 }
663
664 if (vol->reserved_pebs < 0 || vol->alignment < 0 || vol->data_pad < 0 ||
665 vol->name_len < 0) {
666 ubi_err("negative values");
667 goto fail;
668 }
669 if (vol->alignment > ubi->leb_size || vol->alignment == 0) {
670 ubi_err("bad alignment");
671 goto fail;
672 }
673
674 n = vol->alignment % ubi->min_io_size;
675 if (vol->alignment != 1 && n) {
676 ubi_err("alignment is not multiple of min I/O unit");
677 goto fail;
678 }
679
680 n = ubi->leb_size % vol->alignment;
681 if (vol->data_pad != n) {
682 ubi_err("bad data_pad, has to be %lld", n);
683 goto fail;
684 }
685
686 if (vol->vol_type != UBI_DYNAMIC_VOLUME &&
687 vol->vol_type != UBI_STATIC_VOLUME) {
688 ubi_err("bad vol_type");
689 goto fail;
690 }
691
692 if (vol->upd_marker != 0 && vol->upd_marker != 1) {
693 ubi_err("bad upd_marker");
694 goto fail;
695 }
696
697 if (vol->upd_marker && vol->corrupted) {
698 dbg_err("update marker and corrupted simultaneously");
699 goto fail;
700 }
701
702 if (vol->reserved_pebs > ubi->good_peb_count) {
703 ubi_err("too large reserved_pebs");
704 goto fail;
705 }
706
707 n = ubi->leb_size - vol->data_pad;
708 if (vol->usable_leb_size != ubi->leb_size - vol->data_pad) {
709 ubi_err("bad usable_leb_size, has to be %lld", n);
710 goto fail;
711 }
712
713 if (vol->name_len > UBI_VOL_NAME_MAX) {
714 ubi_err("too long volume name, max is %d", UBI_VOL_NAME_MAX);
715 goto fail;
716 }
717
718 if (!vol->name) {
719 ubi_err("NULL volume name");
720 goto fail;
721 }
722
723 n = strnlen(vol->name, vol->name_len + 1);
724 if (n != vol->name_len) {
725 ubi_err("bad name_len %lld", n);
726 goto fail;
727 }
728
729 n = vol->used_ebs * vol->usable_leb_size;
730 if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
731 if (vol->corrupted != 0) {
732 ubi_err("corrupted dynamic volume");
733 goto fail;
734 }
735 if (vol->used_ebs != vol->reserved_pebs) {
736 ubi_err("bad used_ebs");
737 goto fail;
738 }
739 if (vol->last_eb_bytes != vol->usable_leb_size) {
740 ubi_err("bad last_eb_bytes");
741 goto fail;
742 }
743 if (vol->used_bytes != n) {
744 ubi_err("bad used_bytes");
745 goto fail;
746 }
747 } else {
748 if (vol->corrupted != 0 && vol->corrupted != 1) {
749 ubi_err("bad corrupted");
750 goto fail;
751 }
752 if (vol->used_ebs < 0 || vol->used_ebs > vol->reserved_pebs) {
753 ubi_err("bad used_ebs");
754 goto fail;
755 }
756 if (vol->last_eb_bytes < 0 ||
757 vol->last_eb_bytes > vol->usable_leb_size) {
758 ubi_err("bad last_eb_bytes");
759 goto fail;
760 }
761 if (vol->used_bytes < 0 || vol->used_bytes > n ||
762 vol->used_bytes < n - vol->usable_leb_size) {
763 ubi_err("bad used_bytes");
764 goto fail;
765 }
766 }
767
768 alignment = ubi32_to_cpu(ubi->vtbl[vol_id].alignment);
769 data_pad = ubi32_to_cpu(ubi->vtbl[vol_id].data_pad);
770 name_len = ubi16_to_cpu(ubi->vtbl[vol_id].name_len);
771 upd_marker = ubi->vtbl[vol_id].upd_marker;
772 name = &ubi->vtbl[vol_id].name[0];
773 if (ubi->vtbl[vol_id].vol_type == UBI_VID_DYNAMIC)
774 vol_type = UBI_DYNAMIC_VOLUME;
775 else
776 vol_type = UBI_STATIC_VOLUME;
777
778 if (alignment != vol->alignment || data_pad != vol->data_pad ||
779 upd_marker != vol->upd_marker || vol_type != vol->vol_type ||
780 name_len!= vol->name_len || strncmp(name, vol->name, name_len)) {
781 ubi_err("volume info is different");
782 goto fail;
783 }
784
785 return;
786
787fail:
788 ubi_err("paranoid check failed");
789 ubi_dbg_dump_vol_info(vol);
790 ubi_dbg_dump_vtbl_record(&ubi->vtbl[vol_id], vol_id);
791 BUG();
792}
793
794/**
795 * paranoid_check_volumes - check information about all volumes.
796 * @ubi: UBI device description object
797 */
798static void paranoid_check_volumes(struct ubi_device *ubi)
799{
800 int i;
801
802 mutex_lock(&ubi->vtbl_mutex);
803 spin_lock(&ubi->volumes_lock);
804 for (i = 0; i < ubi->vtbl_slots; i++)
805 paranoid_check_volume(ubi, i);
806 spin_unlock(&ubi->volumes_lock);
807 mutex_unlock(&ubi->vtbl_mutex);
808}
809#endif
diff --git a/drivers/mtd/ubi/vtbl.c b/drivers/mtd/ubi/vtbl.c
new file mode 100644
index 000000000000..b6fd6bbd941e
--- /dev/null
+++ b/drivers/mtd/ubi/vtbl.c
@@ -0,0 +1,809 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
3 * Copyright (c) Nokia Corporation, 2006, 2007
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 * the GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 *
19 * Author: Artem Bityutskiy (Битюцкий Артём)
20 */
21
22/*
23 * This file includes volume table manipulation code. The volume table is an
24 * on-flash table containing volume meta-data like name, number of reserved
25 * physical eraseblocks, type, etc. The volume table is stored in the so-called
26 * "layout volume".
27 *
28 * The layout volume is an internal volume which is organized as follows. It
29 * consists of two logical eraseblocks - LEB 0 and LEB 1. Each logical
30 * eraseblock stores one volume table copy, i.e. LEB 0 and LEB 1 duplicate each
31 * other. This redundancy guarantees robustness to unclean reboots. The volume
32 * table is basically an array of volume table records. Each record contains
33 * full information about the volume and protected by a CRC checksum.
34 *
35 * The volume table is changed, it is first changed in RAM. Then LEB 0 is
36 * erased, and the updated volume table is written back to LEB 0. Then same for
37 * LEB 1. This scheme guarantees recoverability from unclean reboots.
38 *
39 * In this UBI implementation the on-flash volume table does not contain any
40 * information about how many data static volumes contain. This information may
41 * be found from the scanning data.
42 *
43 * But it would still be beneficial to store this information in the volume
44 * table. For example, suppose we have a static volume X, and all its physical
45 * eraseblocks became bad for some reasons. Suppose we are attaching the
46 * corresponding MTD device, the scanning has found no logical eraseblocks
47 * corresponding to the volume X. According to the volume table volume X does
48 * exist. So we don't know whether it is just empty or all its physical
49 * eraseblocks went bad. So we cannot alarm the user about this corruption.
50 *
51 * The volume table also stores so-called "update marker", which is used for
52 * volume updates. Before updating the volume, the update marker is set, and
53 * after the update operation is finished, the update marker is cleared. So if
54 * the update operation was interrupted (e.g. by an unclean reboot) - the
55 * update marker is still there and we know that the volume's contents is
56 * damaged.
57 */
58
59#include <linux/crc32.h>
60#include <linux/err.h>
61#include <asm/div64.h>
62#include "ubi.h"
63
64#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
65static void paranoid_vtbl_check(const struct ubi_device *ubi);
66#else
67#define paranoid_vtbl_check(ubi)
68#endif
69
70/* Empty volume table record */
71static struct ubi_vtbl_record empty_vtbl_record;
72
73/**
74 * ubi_change_vtbl_record - change volume table record.
75 * @ubi: UBI device description object
76 * @idx: table index to change
77 * @vtbl_rec: new volume table record
78 *
79 * This function changes volume table record @idx. If @vtbl_rec is %NULL, empty
80 * volume table record is written. The caller does not have to calculate CRC of
81 * the record as it is done by this function. Returns zero in case of success
82 * and a negative error code in case of failure.
83 */
84int ubi_change_vtbl_record(struct ubi_device *ubi, int idx,
85 struct ubi_vtbl_record *vtbl_rec)
86{
87 int i, err;
88 uint32_t crc;
89
90 ubi_assert(idx >= 0 && idx < ubi->vtbl_slots);
91
92 if (!vtbl_rec)
93 vtbl_rec = &empty_vtbl_record;
94 else {
95 crc = crc32(UBI_CRC32_INIT, vtbl_rec, UBI_VTBL_RECORD_SIZE_CRC);
96 vtbl_rec->crc = cpu_to_ubi32(crc);
97 }
98
99 dbg_msg("change record %d", idx);
100 ubi_dbg_dump_vtbl_record(vtbl_rec, idx);
101
102 mutex_lock(&ubi->vtbl_mutex);
103 memcpy(&ubi->vtbl[idx], vtbl_rec, sizeof(struct ubi_vtbl_record));
104 for (i = 0; i < UBI_LAYOUT_VOLUME_EBS; i++) {
105 err = ubi_eba_unmap_leb(ubi, UBI_LAYOUT_VOL_ID, i);
106 if (err) {
107 mutex_unlock(&ubi->vtbl_mutex);
108 return err;
109 }
110 err = ubi_eba_write_leb(ubi, UBI_LAYOUT_VOL_ID, i, ubi->vtbl, 0,
111 ubi->vtbl_size, UBI_LONGTERM);
112 if (err) {
113 mutex_unlock(&ubi->vtbl_mutex);
114 return err;
115 }
116 }
117
118 paranoid_vtbl_check(ubi);
119 mutex_unlock(&ubi->vtbl_mutex);
120 return ubi_wl_flush(ubi);
121}
122
123/**
124 * vol_til_check - check if volume table is not corrupted and contains sensible
125 * data.
126 *
127 * @ubi: UBI device description object
128 * @vtbl: volume table
129 *
130 * This function returns zero if @vtbl is all right, %1 if CRC is incorrect,
131 * and %-EINVAL if it contains inconsistent data.
132 */
133static int vtbl_check(const struct ubi_device *ubi,
134 const struct ubi_vtbl_record *vtbl)
135{
136 int i, n, reserved_pebs, alignment, data_pad, vol_type, name_len;
137 int upd_marker;
138 uint32_t crc;
139 const char *name;
140
141 for (i = 0; i < ubi->vtbl_slots; i++) {
142 cond_resched();
143
144 reserved_pebs = ubi32_to_cpu(vtbl[i].reserved_pebs);
145 alignment = ubi32_to_cpu(vtbl[i].alignment);
146 data_pad = ubi32_to_cpu(vtbl[i].data_pad);
147 upd_marker = vtbl[i].upd_marker;
148 vol_type = vtbl[i].vol_type;
149 name_len = ubi16_to_cpu(vtbl[i].name_len);
150 name = &vtbl[i].name[0];
151
152 crc = crc32(UBI_CRC32_INIT, &vtbl[i], UBI_VTBL_RECORD_SIZE_CRC);
153 if (ubi32_to_cpu(vtbl[i].crc) != crc) {
154 ubi_err("bad CRC at record %u: %#08x, not %#08x",
155 i, crc, ubi32_to_cpu(vtbl[i].crc));
156 ubi_dbg_dump_vtbl_record(&vtbl[i], i);
157 return 1;
158 }
159
160 if (reserved_pebs == 0) {
161 if (memcmp(&vtbl[i], &empty_vtbl_record,
162 UBI_VTBL_RECORD_SIZE)) {
163 dbg_err("bad empty record");
164 goto bad;
165 }
166 continue;
167 }
168
169 if (reserved_pebs < 0 || alignment < 0 || data_pad < 0 ||
170 name_len < 0) {
171 dbg_err("negative values");
172 goto bad;
173 }
174
175 if (alignment > ubi->leb_size || alignment == 0) {
176 dbg_err("bad alignment");
177 goto bad;
178 }
179
180 n = alignment % ubi->min_io_size;
181 if (alignment != 1 && n) {
182 dbg_err("alignment is not multiple of min I/O unit");
183 goto bad;
184 }
185
186 n = ubi->leb_size % alignment;
187 if (data_pad != n) {
188 dbg_err("bad data_pad, has to be %d", n);
189 goto bad;
190 }
191
192 if (vol_type != UBI_VID_DYNAMIC && vol_type != UBI_VID_STATIC) {
193 dbg_err("bad vol_type");
194 goto bad;
195 }
196
197 if (upd_marker != 0 && upd_marker != 1) {
198 dbg_err("bad upd_marker");
199 goto bad;
200 }
201
202 if (reserved_pebs > ubi->good_peb_count) {
203 dbg_err("too large reserved_pebs, good PEBs %d",
204 ubi->good_peb_count);
205 goto bad;
206 }
207
208 if (name_len > UBI_VOL_NAME_MAX) {
209 dbg_err("too long volume name, max %d",
210 UBI_VOL_NAME_MAX);
211 goto bad;
212 }
213
214 if (name[0] == '\0') {
215 dbg_err("NULL volume name");
216 goto bad;
217 }
218
219 if (name_len != strnlen(name, name_len + 1)) {
220 dbg_err("bad name_len");
221 goto bad;
222 }
223 }
224
225 /* Checks that all names are unique */
226 for (i = 0; i < ubi->vtbl_slots - 1; i++) {
227 for (n = i + 1; n < ubi->vtbl_slots; n++) {
228 int len1 = ubi16_to_cpu(vtbl[i].name_len);
229 int len2 = ubi16_to_cpu(vtbl[n].name_len);
230
231 if (len1 > 0 && len1 == len2 &&
232 !strncmp(vtbl[i].name, vtbl[n].name, len1)) {
233 ubi_err("volumes %d and %d have the same name"
234 " \"%s\"", i, n, vtbl[i].name);
235 ubi_dbg_dump_vtbl_record(&vtbl[i], i);
236 ubi_dbg_dump_vtbl_record(&vtbl[n], n);
237 return -EINVAL;
238 }
239 }
240 }
241
242 return 0;
243
244bad:
245 ubi_err("volume table check failed, record %d", i);
246 ubi_dbg_dump_vtbl_record(&vtbl[i], i);
247 return -EINVAL;
248}
249
250/**
251 * create_vtbl - create a copy of volume table.
252 * @ubi: UBI device description object
253 * @si: scanning information
254 * @copy: number of the volume table copy
255 * @vtbl: contents of the volume table
256 *
257 * This function returns zero in case of success and a negative error code in
258 * case of failure.
259 */
260static int create_vtbl(const struct ubi_device *ubi, struct ubi_scan_info *si,
261 int copy, void *vtbl)
262{
263 int err, tries = 0;
264 static struct ubi_vid_hdr *vid_hdr;
265 struct ubi_scan_volume *sv;
266 struct ubi_scan_leb *new_seb, *old_seb = NULL;
267
268 ubi_msg("create volume table (copy #%d)", copy + 1);
269
270 vid_hdr = ubi_zalloc_vid_hdr(ubi);
271 if (!vid_hdr)
272 return -ENOMEM;
273
274 /*
275 * Check if there is a logical eraseblock which would have to contain
276 * this volume table copy was found during scanning. It has to be wiped
277 * out.
278 */
279 sv = ubi_scan_find_sv(si, UBI_LAYOUT_VOL_ID);
280 if (sv)
281 old_seb = ubi_scan_find_seb(sv, copy);
282
283retry:
284 new_seb = ubi_scan_get_free_peb(ubi, si);
285 if (IS_ERR(new_seb)) {
286 err = PTR_ERR(new_seb);
287 goto out_free;
288 }
289
290 vid_hdr->vol_type = UBI_VID_DYNAMIC;
291 vid_hdr->vol_id = cpu_to_ubi32(UBI_LAYOUT_VOL_ID);
292 vid_hdr->compat = UBI_LAYOUT_VOLUME_COMPAT;
293 vid_hdr->data_size = vid_hdr->used_ebs =
294 vid_hdr->data_pad = cpu_to_ubi32(0);
295 vid_hdr->lnum = cpu_to_ubi32(copy);
296 vid_hdr->sqnum = cpu_to_ubi64(++si->max_sqnum);
297 vid_hdr->leb_ver = cpu_to_ubi32(old_seb ? old_seb->leb_ver + 1: 0);
298
299 /* The EC header is already there, write the VID header */
300 err = ubi_io_write_vid_hdr(ubi, new_seb->pnum, vid_hdr);
301 if (err)
302 goto write_error;
303
304 /* Write the layout volume contents */
305 err = ubi_io_write_data(ubi, vtbl, new_seb->pnum, 0, ubi->vtbl_size);
306 if (err)
307 goto write_error;
308
309 /*
310 * And add it to the scanning information. Don't delete the old
311 * @old_seb as it will be deleted and freed in 'ubi_scan_add_used()'.
312 */
313 err = ubi_scan_add_used(ubi, si, new_seb->pnum, new_seb->ec,
314 vid_hdr, 0);
315 kfree(new_seb);
316 ubi_free_vid_hdr(ubi, vid_hdr);
317 return err;
318
319write_error:
320 kfree(new_seb);
321 /* May be this physical eraseblock went bad, try to pick another one */
322 if (++tries <= 5) {
323 err = ubi_scan_add_to_list(si, new_seb->pnum, new_seb->ec,
324 &si->corr);
325 if (!err)
326 goto retry;
327 }
328out_free:
329 ubi_free_vid_hdr(ubi, vid_hdr);
330 return err;
331
332}
333
334/**
335 * process_lvol - process the layout volume.
336 * @ubi: UBI device description object
337 * @si: scanning information
338 * @sv: layout volume scanning information
339 *
340 * This function is responsible for reading the layout volume, ensuring it is
341 * not corrupted, and recovering from corruptions if needed. Returns volume
342 * table in case of success and a negative error code in case of failure.
343 */
344static struct ubi_vtbl_record *process_lvol(const struct ubi_device *ubi,
345 struct ubi_scan_info *si,
346 struct ubi_scan_volume *sv)
347{
348 int err;
349 struct rb_node *rb;
350 struct ubi_scan_leb *seb;
351 struct ubi_vtbl_record *leb[UBI_LAYOUT_VOLUME_EBS] = { NULL, NULL };
352 int leb_corrupted[UBI_LAYOUT_VOLUME_EBS] = {1, 1};
353
354 /*
355 * UBI goes through the following steps when it changes the layout
356 * volume:
357 * a. erase LEB 0;
358 * b. write new data to LEB 0;
359 * c. erase LEB 1;
360 * d. write new data to LEB 1.
361 *
362 * Before the change, both LEBs contain the same data.
363 *
364 * Due to unclean reboots, the contents of LEB 0 may be lost, but there
365 * should LEB 1. So it is OK if LEB 0 is corrupted while LEB 1 is not.
366 * Similarly, LEB 1 may be lost, but there should be LEB 0. And
367 * finally, unclean reboots may result in a situation when neither LEB
368 * 0 nor LEB 1 are corrupted, but they are different. In this case, LEB
369 * 0 contains more recent information.
370 *
371 * So the plan is to first check LEB 0. Then
372 * a. if LEB 0 is OK, it must be containing the most resent data; then
373 * we compare it with LEB 1, and if they are different, we copy LEB
374 * 0 to LEB 1;
375 * b. if LEB 0 is corrupted, but LEB 1 has to be OK, and we copy LEB 1
376 * to LEB 0.
377 */
378
379 dbg_msg("check layout volume");
380
381 /* Read both LEB 0 and LEB 1 into memory */
382 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
383 leb[seb->lnum] = kzalloc(ubi->vtbl_size, GFP_KERNEL);
384 if (!leb[seb->lnum]) {
385 err = -ENOMEM;
386 goto out_free;
387 }
388
389 err = ubi_io_read_data(ubi, leb[seb->lnum], seb->pnum, 0,
390 ubi->vtbl_size);
391 if (err == UBI_IO_BITFLIPS || err == -EBADMSG)
392 /* Scrub the PEB later */
393 seb->scrub = 1;
394 else if (err)
395 goto out_free;
396 }
397
398 err = -EINVAL;
399 if (leb[0]) {
400 leb_corrupted[0] = vtbl_check(ubi, leb[0]);
401 if (leb_corrupted[0] < 0)
402 goto out_free;
403 }
404
405 if (!leb_corrupted[0]) {
406 /* LEB 0 is OK */
407 if (leb[1])
408 leb_corrupted[1] = memcmp(leb[0], leb[1], ubi->vtbl_size);
409 if (leb_corrupted[1]) {
410 ubi_warn("volume table copy #2 is corrupted");
411 err = create_vtbl(ubi, si, 1, leb[0]);
412 if (err)
413 goto out_free;
414 ubi_msg("volume table was restored");
415 }
416
417 /* Both LEB 1 and LEB 2 are OK and consistent */
418 kfree(leb[1]);
419 return leb[0];
420 } else {
421 /* LEB 0 is corrupted or does not exist */
422 if (leb[1]) {
423 leb_corrupted[1] = vtbl_check(ubi, leb[1]);
424 if (leb_corrupted[1] < 0)
425 goto out_free;
426 }
427 if (leb_corrupted[1]) {
428 /* Both LEB 0 and LEB 1 are corrupted */
429 ubi_err("both volume tables are corrupted");
430 goto out_free;
431 }
432
433 ubi_warn("volume table copy #1 is corrupted");
434 err = create_vtbl(ubi, si, 0, leb[1]);
435 if (err)
436 goto out_free;
437 ubi_msg("volume table was restored");
438
439 kfree(leb[0]);
440 return leb[1];
441 }
442
443out_free:
444 kfree(leb[0]);
445 kfree(leb[1]);
446 return ERR_PTR(err);
447}
448
449/**
450 * create_empty_lvol - create empty layout volume.
451 * @ubi: UBI device description object
452 * @si: scanning information
453 *
454 * This function returns volume table contents in case of success and a
455 * negative error code in case of failure.
456 */
457static struct ubi_vtbl_record *create_empty_lvol(const struct ubi_device *ubi,
458 struct ubi_scan_info *si)
459{
460 int i;
461 struct ubi_vtbl_record *vtbl;
462
463 vtbl = kzalloc(ubi->vtbl_size, GFP_KERNEL);
464 if (!vtbl)
465 return ERR_PTR(-ENOMEM);
466
467 for (i = 0; i < ubi->vtbl_slots; i++)
468 memcpy(&vtbl[i], &empty_vtbl_record, UBI_VTBL_RECORD_SIZE);
469
470 for (i = 0; i < UBI_LAYOUT_VOLUME_EBS; i++) {
471 int err;
472
473 err = create_vtbl(ubi, si, i, vtbl);
474 if (err) {
475 kfree(vtbl);
476 return ERR_PTR(err);
477 }
478 }
479
480 return vtbl;
481}
482
483/**
484 * init_volumes - initialize volume information for existing volumes.
485 * @ubi: UBI device description object
486 * @si: scanning information
487 * @vtbl: volume table
488 *
489 * This function allocates volume description objects for existing volumes.
490 * Returns zero in case of success and a negative error code in case of
491 * failure.
492 */
493static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
494 const struct ubi_vtbl_record *vtbl)
495{
496 int i, reserved_pebs = 0;
497 struct ubi_scan_volume *sv;
498 struct ubi_volume *vol;
499
500 for (i = 0; i < ubi->vtbl_slots; i++) {
501 cond_resched();
502
503 if (ubi32_to_cpu(vtbl[i].reserved_pebs) == 0)
504 continue; /* Empty record */
505
506 vol = kzalloc(sizeof(struct ubi_volume), GFP_KERNEL);
507 if (!vol)
508 return -ENOMEM;
509
510 vol->reserved_pebs = ubi32_to_cpu(vtbl[i].reserved_pebs);
511 vol->alignment = ubi32_to_cpu(vtbl[i].alignment);
512 vol->data_pad = ubi32_to_cpu(vtbl[i].data_pad);
513 vol->vol_type = vtbl[i].vol_type == UBI_VID_DYNAMIC ?
514 UBI_DYNAMIC_VOLUME : UBI_STATIC_VOLUME;
515 vol->name_len = ubi16_to_cpu(vtbl[i].name_len);
516 vol->usable_leb_size = ubi->leb_size - vol->data_pad;
517 memcpy(vol->name, vtbl[i].name, vol->name_len);
518 vol->name[vol->name_len] = '\0';
519 vol->vol_id = i;
520
521 ubi_assert(!ubi->volumes[i]);
522 ubi->volumes[i] = vol;
523 ubi->vol_count += 1;
524 vol->ubi = ubi;
525 reserved_pebs += vol->reserved_pebs;
526
527 /*
528 * In case of dynamic volume UBI knows nothing about how many
529 * data is stored there. So assume the whole volume is used.
530 */
531 if (vol->vol_type == UBI_DYNAMIC_VOLUME) {
532 vol->used_ebs = vol->reserved_pebs;
533 vol->last_eb_bytes = vol->usable_leb_size;
534 vol->used_bytes = vol->used_ebs * vol->usable_leb_size;
535 continue;
536 }
537
538 /* Static volumes only */
539 sv = ubi_scan_find_sv(si, i);
540 if (!sv) {
541 /*
542 * No eraseblocks belonging to this volume found. We
543 * don't actually know whether this static volume is
544 * completely corrupted or just contains no data. And
545 * we cannot know this as long as data size is not
546 * stored on flash. So we just assume the volume is
547 * empty. FIXME: this should be handled.
548 */
549 continue;
550 }
551
552 if (sv->leb_count != sv->used_ebs) {
553 /*
554 * We found a static volume which misses several
555 * eraseblocks. Treat it as corrupted.
556 */
557 ubi_warn("static volume %d misses %d LEBs - corrupted",
558 sv->vol_id, sv->used_ebs - sv->leb_count);
559 vol->corrupted = 1;
560 continue;
561 }
562
563 vol->used_ebs = sv->used_ebs;
564 vol->used_bytes = (vol->used_ebs - 1) * vol->usable_leb_size;
565 vol->used_bytes += sv->last_data_size;
566 vol->last_eb_bytes = sv->last_data_size;
567 }
568
569 vol = kzalloc(sizeof(struct ubi_volume), GFP_KERNEL);
570 if (!vol)
571 return -ENOMEM;
572
573 vol->reserved_pebs = UBI_LAYOUT_VOLUME_EBS;
574 vol->alignment = 1;
575 vol->vol_type = UBI_DYNAMIC_VOLUME;
576 vol->name_len = sizeof(UBI_LAYOUT_VOLUME_NAME) - 1;
577 memcpy(vol->name, UBI_LAYOUT_VOLUME_NAME, vol->name_len + 1);
578 vol->usable_leb_size = ubi->leb_size;
579 vol->used_ebs = vol->reserved_pebs;
580 vol->last_eb_bytes = vol->reserved_pebs;
581 vol->used_bytes = vol->used_ebs * (ubi->leb_size - vol->data_pad);
582 vol->vol_id = UBI_LAYOUT_VOL_ID;
583
584 ubi_assert(!ubi->volumes[i]);
585 ubi->volumes[vol_id2idx(ubi, vol->vol_id)] = vol;
586 reserved_pebs += vol->reserved_pebs;
587 ubi->vol_count += 1;
588 vol->ubi = ubi;
589
590 if (reserved_pebs > ubi->avail_pebs)
591 ubi_err("not enough PEBs, required %d, available %d",
592 reserved_pebs, ubi->avail_pebs);
593 ubi->rsvd_pebs += reserved_pebs;
594 ubi->avail_pebs -= reserved_pebs;
595
596 return 0;
597}
598
599/**
600 * check_sv - check volume scanning information.
601 * @vol: UBI volume description object
602 * @sv: volume scanning information
603 *
604 * This function returns zero if the volume scanning information is consistent
605 * to the data read from the volume tabla, and %-EINVAL if not.
606 */
607static int check_sv(const struct ubi_volume *vol,
608 const struct ubi_scan_volume *sv)
609{
610 if (sv->highest_lnum >= vol->reserved_pebs) {
611 dbg_err("bad highest_lnum");
612 goto bad;
613 }
614 if (sv->leb_count > vol->reserved_pebs) {
615 dbg_err("bad leb_count");
616 goto bad;
617 }
618 if (sv->vol_type != vol->vol_type) {
619 dbg_err("bad vol_type");
620 goto bad;
621 }
622 if (sv->used_ebs > vol->reserved_pebs) {
623 dbg_err("bad used_ebs");
624 goto bad;
625 }
626 if (sv->data_pad != vol->data_pad) {
627 dbg_err("bad data_pad");
628 goto bad;
629 }
630 return 0;
631
632bad:
633 ubi_err("bad scanning information");
634 ubi_dbg_dump_sv(sv);
635 ubi_dbg_dump_vol_info(vol);
636 return -EINVAL;
637}
638
639/**
640 * check_scanning_info - check that scanning information.
641 * @ubi: UBI device description object
642 * @si: scanning information
643 *
644 * Even though we protect on-flash data by CRC checksums, we still don't trust
645 * the media. This function ensures that scanning information is consistent to
646 * the information read from the volume table. Returns zero if the scanning
647 * information is OK and %-EINVAL if it is not.
648 */
649static int check_scanning_info(const struct ubi_device *ubi,
650 struct ubi_scan_info *si)
651{
652 int err, i;
653 struct ubi_scan_volume *sv;
654 struct ubi_volume *vol;
655
656 if (si->vols_found > UBI_INT_VOL_COUNT + ubi->vtbl_slots) {
657 ubi_err("scanning found %d volumes, maximum is %d + %d",
658 si->vols_found, UBI_INT_VOL_COUNT, ubi->vtbl_slots);
659 return -EINVAL;
660 }
661
662 if (si->highest_vol_id >= ubi->vtbl_slots + UBI_INT_VOL_COUNT&&
663 si->highest_vol_id < UBI_INTERNAL_VOL_START) {
664 ubi_err("too large volume ID %d found by scanning",
665 si->highest_vol_id);
666 return -EINVAL;
667 }
668
669
670 for (i = 0; i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++) {
671 cond_resched();
672
673 sv = ubi_scan_find_sv(si, i);
674 vol = ubi->volumes[i];
675 if (!vol) {
676 if (sv)
677 ubi_scan_rm_volume(si, sv);
678 continue;
679 }
680
681 if (vol->reserved_pebs == 0) {
682 ubi_assert(i < ubi->vtbl_slots);
683
684 if (!sv)
685 continue;
686
687 /*
688 * During scanning we found a volume which does not
689 * exist according to the information in the volume
690 * table. This must have happened due to an unclean
691 * reboot while the volume was being removed. Discard
692 * these eraseblocks.
693 */
694 ubi_msg("finish volume %d removal", sv->vol_id);
695 ubi_scan_rm_volume(si, sv);
696 } else if (sv) {
697 err = check_sv(vol, sv);
698 if (err)
699 return err;
700 }
701 }
702
703 return 0;
704}
705
706/**
707 * ubi_read_volume_table - read volume table.
708 * information.
709 * @ubi: UBI device description object
710 * @si: scanning information
711 *
712 * This function reads volume table, checks it, recover from errors if needed,
713 * or creates it if needed. Returns zero in case of success and a negative
714 * error code in case of failure.
715 */
716int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si)
717{
718 int i, err;
719 struct ubi_scan_volume *sv;
720
721 empty_vtbl_record.crc = cpu_to_ubi32(0xf116c36b);
722
723 /*
724 * The number of supported volumes is limited by the eraseblock size
725 * and by the UBI_MAX_VOLUMES constant.
726 */
727 ubi->vtbl_slots = ubi->leb_size / UBI_VTBL_RECORD_SIZE;
728 if (ubi->vtbl_slots > UBI_MAX_VOLUMES)
729 ubi->vtbl_slots = UBI_MAX_VOLUMES;
730
731 ubi->vtbl_size = ubi->vtbl_slots * UBI_VTBL_RECORD_SIZE;
732 ubi->vtbl_size = ALIGN(ubi->vtbl_size, ubi->min_io_size);
733
734 sv = ubi_scan_find_sv(si, UBI_LAYOUT_VOL_ID);
735 if (!sv) {
736 /*
737 * No logical eraseblocks belonging to the layout volume were
738 * found. This could mean that the flash is just empty. In
739 * this case we create empty layout volume.
740 *
741 * But if flash is not empty this must be a corruption or the
742 * MTD device just contains garbage.
743 */
744 if (si->is_empty) {
745 ubi->vtbl = create_empty_lvol(ubi, si);
746 if (IS_ERR(ubi->vtbl))
747 return PTR_ERR(ubi->vtbl);
748 } else {
749 ubi_err("the layout volume was not found");
750 return -EINVAL;
751 }
752 } else {
753 if (sv->leb_count > UBI_LAYOUT_VOLUME_EBS) {
754 /* This must not happen with proper UBI images */
755 dbg_err("too many LEBs (%d) in layout volume",
756 sv->leb_count);
757 return -EINVAL;
758 }
759
760 ubi->vtbl = process_lvol(ubi, si, sv);
761 if (IS_ERR(ubi->vtbl))
762 return PTR_ERR(ubi->vtbl);
763 }
764
765 ubi->avail_pebs = ubi->good_peb_count;
766
767 /*
768 * The layout volume is OK, initialize the corresponding in-RAM data
769 * structures.
770 */
771 err = init_volumes(ubi, si, ubi->vtbl);
772 if (err)
773 goto out_free;
774
775 /*
776 * Get sure that the scanning information is consistent to the
777 * information stored in the volume table.
778 */
779 err = check_scanning_info(ubi, si);
780 if (err)
781 goto out_free;
782
783 return 0;
784
785out_free:
786 kfree(ubi->vtbl);
787 for (i = 0; i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++)
788 if (ubi->volumes[i]) {
789 kfree(ubi->volumes[i]);
790 ubi->volumes[i] = NULL;
791 }
792 return err;
793}
794
795#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
796
797/**
798 * paranoid_vtbl_check - check volume table.
799 * @ubi: UBI device description object
800 */
801static void paranoid_vtbl_check(const struct ubi_device *ubi)
802{
803 if (vtbl_check(ubi, ubi->vtbl)) {
804 ubi_err("paranoid check failed");
805 BUG();
806 }
807}
808
809#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
new file mode 100644
index 000000000000..9ecaf77eca9e
--- /dev/null
+++ b/drivers/mtd/ubi/wl.c
@@ -0,0 +1,1671 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Authors: Artem Bityutskiy (Битюцкий Артём), Thomas Gleixner
19 */
20
21/*
22 * UBI wear-leveling unit.
23 *
24 * This unit is responsible for wear-leveling. It works in terms of physical
25 * eraseblocks and erase counters and knows nothing about logical eraseblocks,
26 * volumes, etc. From this unit's perspective all physical eraseblocks are of
27 * two types - used and free. Used physical eraseblocks are those that were
28 * "get" by the 'ubi_wl_get_peb()' function, and free physical eraseblocks are
29 * those that were put by the 'ubi_wl_put_peb()' function.
30 *
31 * Physical eraseblocks returned by 'ubi_wl_get_peb()' have only erase counter
32 * header. The rest of the physical eraseblock contains only 0xFF bytes.
33 *
34 * When physical eraseblocks are returned to the WL unit by means of the
35 * 'ubi_wl_put_peb()' function, they are scheduled for erasure. The erasure is
36 * done asynchronously in context of the per-UBI device background thread,
37 * which is also managed by the WL unit.
38 *
39 * The wear-leveling is ensured by means of moving the contents of used
40 * physical eraseblocks with low erase counter to free physical eraseblocks
41 * with high erase counter.
42 *
43 * The 'ubi_wl_get_peb()' function accepts data type hints which help to pick
44 * an "optimal" physical eraseblock. For example, when it is known that the
45 * physical eraseblock will be "put" soon because it contains short-term data,
46 * the WL unit may pick a free physical eraseblock with low erase counter, and
47 * so forth.
48 *
49 * If the WL unit fails to erase a physical eraseblock, it marks it as bad.
50 *
51 * This unit is also responsible for scrubbing. If a bit-flip is detected in a
52 * physical eraseblock, it has to be moved. Technically this is the same as
53 * moving it for wear-leveling reasons.
54 *
55 * As it was said, for the UBI unit all physical eraseblocks are either "free"
56 * or "used". Free eraseblock are kept in the @wl->free RB-tree, while used
57 * eraseblocks are kept in a set of different RB-trees: @wl->used,
58 * @wl->prot.pnum, @wl->prot.aec, and @wl->scrub.
59 *
60 * Note, in this implementation, we keep a small in-RAM object for each physical
61 * eraseblock. This is surely not a scalable solution. But it appears to be good
62 * enough for moderately large flashes and it is simple. In future, one may
63 * re-work this unit and make it more scalable.
64 *
65 * At the moment this unit does not utilize the sequence number, which was
66 * introduced relatively recently. But it would be wise to do this because the
67 * sequence number of a logical eraseblock characterizes how old is it. For
68 * example, when we move a PEB with low erase counter, and we need to pick the
69 * target PEB, we pick a PEB with the highest EC if our PEB is "old" and we
70 * pick target PEB with an average EC if our PEB is not very "old". This is a
71 * room for future re-works of the WL unit.
72 *
73 * FIXME: looks too complex, should be simplified (later).
74 */
75
76#include <linux/slab.h>
77#include <linux/crc32.h>
78#include <linux/freezer.h>
79#include <linux/kthread.h>
80#include "ubi.h"
81
82/* Number of physical eraseblocks reserved for wear-leveling purposes */
83#define WL_RESERVED_PEBS 1
84
85/*
86 * How many erase cycles are short term, unknown, and long term physical
87 * eraseblocks protected.
88 */
89#define ST_PROTECTION 16
90#define U_PROTECTION 10
91#define LT_PROTECTION 4
92
93/*
94 * Maximum difference between two erase counters. If this threshold is
95 * exceeded, the WL unit starts moving data from used physical eraseblocks with
96 * low erase counter to free physical eraseblocks with high erase counter.
97 */
98#define UBI_WL_THRESHOLD CONFIG_MTD_UBI_WL_THRESHOLD
99
100/*
101 * When a physical eraseblock is moved, the WL unit has to pick the target
102 * physical eraseblock to move to. The simplest way would be just to pick the
103 * one with the highest erase counter. But in certain workloads this could lead
104 * to an unlimited wear of one or few physical eraseblock. Indeed, imagine a
105 * situation when the picked physical eraseblock is constantly erased after the
106 * data is written to it. So, we have a constant which limits the highest erase
107 * counter of the free physical eraseblock to pick. Namely, the WL unit does
108 * not pick eraseblocks with erase counter greater then the lowest erase
109 * counter plus %WL_FREE_MAX_DIFF.
110 */
111#define WL_FREE_MAX_DIFF (2*UBI_WL_THRESHOLD)
112
113/*
114 * Maximum number of consecutive background thread failures which is enough to
115 * switch to read-only mode.
116 */
117#define WL_MAX_FAILURES 32
118
119/**
120 * struct ubi_wl_entry - wear-leveling entry.
121 * @rb: link in the corresponding RB-tree
122 * @ec: erase counter
123 * @pnum: physical eraseblock number
124 *
125 * Each physical eraseblock has a corresponding &struct wl_entry object which
126 * may be kept in different RB-trees.
127 */
128struct ubi_wl_entry {
129 struct rb_node rb;
130 int ec;
131 int pnum;
132};
133
134/**
135 * struct ubi_wl_prot_entry - PEB protection entry.
136 * @rb_pnum: link in the @wl->prot.pnum RB-tree
137 * @rb_aec: link in the @wl->prot.aec RB-tree
138 * @abs_ec: the absolute erase counter value when the protection ends
139 * @e: the wear-leveling entry of the physical eraseblock under protection
140 *
141 * When the WL unit returns a physical eraseblock, the physical eraseblock is
142 * protected from being moved for some "time". For this reason, the physical
143 * eraseblock is not directly moved from the @wl->free tree to the @wl->used
144 * tree. There is one more tree in between where this physical eraseblock is
145 * temporarily stored (@wl->prot).
146 *
147 * All this protection stuff is needed because:
148 * o we don't want to move physical eraseblocks just after we have given them
149 * to the user; instead, we first want to let users fill them up with data;
150 *
151 * o there is a chance that the user will put the physical eraseblock very
152 * soon, so it makes sense not to move it for some time, but wait; this is
153 * especially important in case of "short term" physical eraseblocks.
154 *
155 * Physical eraseblocks stay protected only for limited time. But the "time" is
156 * measured in erase cycles in this case. This is implemented with help of the
157 * absolute erase counter (@wl->abs_ec). When it reaches certain value, the
158 * physical eraseblocks are moved from the protection trees (@wl->prot.*) to
159 * the @wl->used tree.
160 *
161 * Protected physical eraseblocks are searched by physical eraseblock number
162 * (when they are put) and by the absolute erase counter (to check if it is
163 * time to move them to the @wl->used tree). So there are actually 2 RB-trees
164 * storing the protected physical eraseblocks: @wl->prot.pnum and
165 * @wl->prot.aec. They are referred to as the "protection" trees. The
166 * first one is indexed by the physical eraseblock number. The second one is
167 * indexed by the absolute erase counter. Both trees store
168 * &struct ubi_wl_prot_entry objects.
169 *
170 * Each physical eraseblock has 2 main states: free and used. The former state
171 * corresponds to the @wl->free tree. The latter state is split up on several
172 * sub-states:
173 * o the WL movement is allowed (@wl->used tree);
174 * o the WL movement is temporarily prohibited (@wl->prot.pnum and
175 * @wl->prot.aec trees);
176 * o scrubbing is needed (@wl->scrub tree).
177 *
178 * Depending on the sub-state, wear-leveling entries of the used physical
179 * eraseblocks may be kept in one of those trees.
180 */
181struct ubi_wl_prot_entry {
182 struct rb_node rb_pnum;
183 struct rb_node rb_aec;
184 unsigned long long abs_ec;
185 struct ubi_wl_entry *e;
186};
187
188/**
189 * struct ubi_work - UBI work description data structure.
190 * @list: a link in the list of pending works
191 * @func: worker function
192 * @priv: private data of the worker function
193 *
194 * @e: physical eraseblock to erase
195 * @torture: if the physical eraseblock has to be tortured
196 *
197 * The @func pointer points to the worker function. If the @cancel argument is
198 * not zero, the worker has to free the resources and exit immediately. The
199 * worker has to return zero in case of success and a negative error code in
200 * case of failure.
201 */
202struct ubi_work {
203 struct list_head list;
204 int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int cancel);
205 /* The below fields are only relevant to erasure works */
206 struct ubi_wl_entry *e;
207 int torture;
208};
209
210#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
211static int paranoid_check_ec(const struct ubi_device *ubi, int pnum, int ec);
212static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e,
213 struct rb_root *root);
214#else
215#define paranoid_check_ec(ubi, pnum, ec) 0
216#define paranoid_check_in_wl_tree(e, root)
217#endif
218
219/* Slab cache for wear-leveling entries */
220static struct kmem_cache *wl_entries_slab;
221
222/**
223 * tree_empty - a helper function to check if an RB-tree is empty.
224 * @root: the root of the tree
225 *
226 * This function returns non-zero if the RB-tree is empty and zero if not.
227 */
228static inline int tree_empty(struct rb_root *root)
229{
230 return root->rb_node == NULL;
231}
232
233/**
234 * wl_tree_add - add a wear-leveling entry to a WL RB-tree.
235 * @e: the wear-leveling entry to add
236 * @root: the root of the tree
237 *
238 * Note, we use (erase counter, physical eraseblock number) pairs as keys in
239 * the @ubi->used and @ubi->free RB-trees.
240 */
241static void wl_tree_add(struct ubi_wl_entry *e, struct rb_root *root)
242{
243 struct rb_node **p, *parent = NULL;
244
245 p = &root->rb_node;
246 while (*p) {
247 struct ubi_wl_entry *e1;
248
249 parent = *p;
250 e1 = rb_entry(parent, struct ubi_wl_entry, rb);
251
252 if (e->ec < e1->ec)
253 p = &(*p)->rb_left;
254 else if (e->ec > e1->ec)
255 p = &(*p)->rb_right;
256 else {
257 ubi_assert(e->pnum != e1->pnum);
258 if (e->pnum < e1->pnum)
259 p = &(*p)->rb_left;
260 else
261 p = &(*p)->rb_right;
262 }
263 }
264
265 rb_link_node(&e->rb, parent, p);
266 rb_insert_color(&e->rb, root);
267}
268
269
270/*
271 * Helper functions to add and delete wear-leveling entries from different
272 * trees.
273 */
274
275static void free_tree_add(struct ubi_device *ubi, struct ubi_wl_entry *e)
276{
277 wl_tree_add(e, &ubi->free);
278}
279static inline void used_tree_add(struct ubi_device *ubi,
280 struct ubi_wl_entry *e)
281{
282 wl_tree_add(e, &ubi->used);
283}
284static inline void scrub_tree_add(struct ubi_device *ubi,
285 struct ubi_wl_entry *e)
286{
287 wl_tree_add(e, &ubi->scrub);
288}
289static inline void free_tree_del(struct ubi_device *ubi,
290 struct ubi_wl_entry *e)
291{
292 paranoid_check_in_wl_tree(e, &ubi->free);
293 rb_erase(&e->rb, &ubi->free);
294}
295static inline void used_tree_del(struct ubi_device *ubi,
296 struct ubi_wl_entry *e)
297{
298 paranoid_check_in_wl_tree(e, &ubi->used);
299 rb_erase(&e->rb, &ubi->used);
300}
301static inline void scrub_tree_del(struct ubi_device *ubi,
302 struct ubi_wl_entry *e)
303{
304 paranoid_check_in_wl_tree(e, &ubi->scrub);
305 rb_erase(&e->rb, &ubi->scrub);
306}
307
308/**
309 * do_work - do one pending work.
310 * @ubi: UBI device description object
311 *
312 * This function returns zero in case of success and a negative error code in
313 * case of failure.
314 */
315static int do_work(struct ubi_device *ubi)
316{
317 int err;
318 struct ubi_work *wrk;
319
320 spin_lock(&ubi->wl_lock);
321
322 if (list_empty(&ubi->works)) {
323 spin_unlock(&ubi->wl_lock);
324 return 0;
325 }
326
327 wrk = list_entry(ubi->works.next, struct ubi_work, list);
328 list_del(&wrk->list);
329 spin_unlock(&ubi->wl_lock);
330
331 /*
332 * Call the worker function. Do not touch the work structure
333 * after this call as it will have been freed or reused by that
334 * time by the worker function.
335 */
336 err = wrk->func(ubi, wrk, 0);
337 if (err)
338 ubi_err("work failed with error code %d", err);
339
340 spin_lock(&ubi->wl_lock);
341 ubi->works_count -= 1;
342 ubi_assert(ubi->works_count >= 0);
343 spin_unlock(&ubi->wl_lock);
344 return err;
345}
346
347/**
348 * produce_free_peb - produce a free physical eraseblock.
349 * @ubi: UBI device description object
350 *
351 * This function tries to make a free PEB by means of synchronous execution of
352 * pending works. This may be needed if, for example the background thread is
353 * disabled. Returns zero in case of success and a negative error code in case
354 * of failure.
355 */
356static int produce_free_peb(struct ubi_device *ubi)
357{
358 int err;
359
360 spin_lock(&ubi->wl_lock);
361 while (tree_empty(&ubi->free)) {
362 spin_unlock(&ubi->wl_lock);
363
364 dbg_wl("do one work synchronously");
365 err = do_work(ubi);
366 if (err)
367 return err;
368
369 spin_lock(&ubi->wl_lock);
370 }
371 spin_unlock(&ubi->wl_lock);
372
373 return 0;
374}
375
376/**
377 * in_wl_tree - check if wear-leveling entry is present in a WL RB-tree.
378 * @e: the wear-leveling entry to check
379 * @root: the root of the tree
380 *
381 * This function returns non-zero if @e is in the @root RB-tree and zero if it
382 * is not.
383 */
384static int in_wl_tree(struct ubi_wl_entry *e, struct rb_root *root)
385{
386 struct rb_node *p;
387
388 p = root->rb_node;
389 while (p) {
390 struct ubi_wl_entry *e1;
391
392 e1 = rb_entry(p, struct ubi_wl_entry, rb);
393
394 if (e->pnum == e1->pnum) {
395 ubi_assert(e == e1);
396 return 1;
397 }
398
399 if (e->ec < e1->ec)
400 p = p->rb_left;
401 else if (e->ec > e1->ec)
402 p = p->rb_right;
403 else {
404 ubi_assert(e->pnum != e1->pnum);
405 if (e->pnum < e1->pnum)
406 p = p->rb_left;
407 else
408 p = p->rb_right;
409 }
410 }
411
412 return 0;
413}
414
415/**
416 * prot_tree_add - add physical eraseblock to protection trees.
417 * @ubi: UBI device description object
418 * @e: the physical eraseblock to add
419 * @pe: protection entry object to use
420 * @abs_ec: absolute erase counter value when this physical eraseblock has
421 * to be removed from the protection trees.
422 *
423 * @wl->lock has to be locked.
424 */
425static void prot_tree_add(struct ubi_device *ubi, struct ubi_wl_entry *e,
426 struct ubi_wl_prot_entry *pe, int abs_ec)
427{
428 struct rb_node **p, *parent = NULL;
429 struct ubi_wl_prot_entry *pe1;
430
431 pe->e = e;
432 pe->abs_ec = ubi->abs_ec + abs_ec;
433
434 p = &ubi->prot.pnum.rb_node;
435 while (*p) {
436 parent = *p;
437 pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_pnum);
438
439 if (e->pnum < pe1->e->pnum)
440 p = &(*p)->rb_left;
441 else
442 p = &(*p)->rb_right;
443 }
444 rb_link_node(&pe->rb_pnum, parent, p);
445 rb_insert_color(&pe->rb_pnum, &ubi->prot.pnum);
446
447 p = &ubi->prot.aec.rb_node;
448 parent = NULL;
449 while (*p) {
450 parent = *p;
451 pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_aec);
452
453 if (pe->abs_ec < pe1->abs_ec)
454 p = &(*p)->rb_left;
455 else
456 p = &(*p)->rb_right;
457 }
458 rb_link_node(&pe->rb_aec, parent, p);
459 rb_insert_color(&pe->rb_aec, &ubi->prot.aec);
460}
461
462/**
463 * find_wl_entry - find wear-leveling entry closest to certain erase counter.
464 * @root: the RB-tree where to look for
465 * @max: highest possible erase counter
466 *
467 * This function looks for a wear leveling entry with erase counter closest to
468 * @max and less then @max.
469 */
470static struct ubi_wl_entry *find_wl_entry(struct rb_root *root, int max)
471{
472 struct rb_node *p;
473 struct ubi_wl_entry *e;
474
475 e = rb_entry(rb_first(root), struct ubi_wl_entry, rb);
476 max += e->ec;
477
478 p = root->rb_node;
479 while (p) {
480 struct ubi_wl_entry *e1;
481
482 e1 = rb_entry(p, struct ubi_wl_entry, rb);
483 if (e1->ec >= max)
484 p = p->rb_left;
485 else {
486 p = p->rb_right;
487 e = e1;
488 }
489 }
490
491 return e;
492}
493
494/**
495 * ubi_wl_get_peb - get a physical eraseblock.
496 * @ubi: UBI device description object
497 * @dtype: type of data which will be stored in this physical eraseblock
498 *
499 * This function returns a physical eraseblock in case of success and a
500 * negative error code in case of failure. Might sleep.
501 */
502int ubi_wl_get_peb(struct ubi_device *ubi, int dtype)
503{
504 int err, protect, medium_ec;
505 struct ubi_wl_entry *e, *first, *last;
506 struct ubi_wl_prot_entry *pe;
507
508 ubi_assert(dtype == UBI_LONGTERM || dtype == UBI_SHORTTERM ||
509 dtype == UBI_UNKNOWN);
510
511 pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_KERNEL);
512 if (!pe)
513 return -ENOMEM;
514
515retry:
516 spin_lock(&ubi->wl_lock);
517 if (tree_empty(&ubi->free)) {
518 if (ubi->works_count == 0) {
519 ubi_assert(list_empty(&ubi->works));
520 ubi_err("no free eraseblocks");
521 spin_unlock(&ubi->wl_lock);
522 kfree(pe);
523 return -ENOSPC;
524 }
525 spin_unlock(&ubi->wl_lock);
526
527 err = produce_free_peb(ubi);
528 if (err < 0) {
529 kfree(pe);
530 return err;
531 }
532 goto retry;
533 }
534
535 switch (dtype) {
536 case UBI_LONGTERM:
537 /*
538 * For long term data we pick a physical eraseblock
539 * with high erase counter. But the highest erase
540 * counter we can pick is bounded by the the lowest
541 * erase counter plus %WL_FREE_MAX_DIFF.
542 */
543 e = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
544 protect = LT_PROTECTION;
545 break;
546 case UBI_UNKNOWN:
547 /*
548 * For unknown data we pick a physical eraseblock with
549 * medium erase counter. But we by no means can pick a
550 * physical eraseblock with erase counter greater or
551 * equivalent than the lowest erase counter plus
552 * %WL_FREE_MAX_DIFF.
553 */
554 first = rb_entry(rb_first(&ubi->free),
555 struct ubi_wl_entry, rb);
556 last = rb_entry(rb_last(&ubi->free),
557 struct ubi_wl_entry, rb);
558
559 if (last->ec - first->ec < WL_FREE_MAX_DIFF)
560 e = rb_entry(ubi->free.rb_node,
561 struct ubi_wl_entry, rb);
562 else {
563 medium_ec = (first->ec + WL_FREE_MAX_DIFF)/2;
564 e = find_wl_entry(&ubi->free, medium_ec);
565 }
566 protect = U_PROTECTION;
567 break;
568 case UBI_SHORTTERM:
569 /*
570 * For short term data we pick a physical eraseblock
571 * with the lowest erase counter as we expect it will
572 * be erased soon.
573 */
574 e = rb_entry(rb_first(&ubi->free),
575 struct ubi_wl_entry, rb);
576 protect = ST_PROTECTION;
577 break;
578 default:
579 protect = 0;
580 e = NULL;
581 BUG();
582 }
583
584 /*
585 * Move the physical eraseblock to the protection trees where it will
586 * be protected from being moved for some time.
587 */
588 free_tree_del(ubi, e);
589 prot_tree_add(ubi, e, pe, protect);
590
591 dbg_wl("PEB %d EC %d, protection %d", e->pnum, e->ec, protect);
592 spin_unlock(&ubi->wl_lock);
593
594 return e->pnum;
595}
596
597/**
598 * prot_tree_del - remove a physical eraseblock from the protection trees
599 * @ubi: UBI device description object
600 * @pnum: the physical eraseblock to remove
601 */
602static void prot_tree_del(struct ubi_device *ubi, int pnum)
603{
604 struct rb_node *p;
605 struct ubi_wl_prot_entry *pe = NULL;
606
607 p = ubi->prot.pnum.rb_node;
608 while (p) {
609
610 pe = rb_entry(p, struct ubi_wl_prot_entry, rb_pnum);
611
612 if (pnum == pe->e->pnum)
613 break;
614
615 if (pnum < pe->e->pnum)
616 p = p->rb_left;
617 else
618 p = p->rb_right;
619 }
620
621 ubi_assert(pe->e->pnum == pnum);
622 rb_erase(&pe->rb_aec, &ubi->prot.aec);
623 rb_erase(&pe->rb_pnum, &ubi->prot.pnum);
624 kfree(pe);
625}
626
627/**
628 * sync_erase - synchronously erase a physical eraseblock.
629 * @ubi: UBI device description object
630 * @e: the the physical eraseblock to erase
631 * @torture: if the physical eraseblock has to be tortured
632 *
633 * This function returns zero in case of success and a negative error code in
634 * case of failure.
635 */
636static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, int torture)
637{
638 int err;
639 struct ubi_ec_hdr *ec_hdr;
640 unsigned long long ec = e->ec;
641
642 dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec);
643
644 err = paranoid_check_ec(ubi, e->pnum, e->ec);
645 if (err > 0)
646 return -EINVAL;
647
648 ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
649 if (!ec_hdr)
650 return -ENOMEM;
651
652 err = ubi_io_sync_erase(ubi, e->pnum, torture);
653 if (err < 0)
654 goto out_free;
655
656 ec += err;
657 if (ec > UBI_MAX_ERASECOUNTER) {
658 /*
659 * Erase counter overflow. Upgrade UBI and use 64-bit
660 * erase counters internally.
661 */
662 ubi_err("erase counter overflow at PEB %d, EC %llu",
663 e->pnum, ec);
664 err = -EINVAL;
665 goto out_free;
666 }
667
668 dbg_wl("erased PEB %d, new EC %llu", e->pnum, ec);
669
670 ec_hdr->ec = cpu_to_ubi64(ec);
671
672 err = ubi_io_write_ec_hdr(ubi, e->pnum, ec_hdr);
673 if (err)
674 goto out_free;
675
676 e->ec = ec;
677 spin_lock(&ubi->wl_lock);
678 if (e->ec > ubi->max_ec)
679 ubi->max_ec = e->ec;
680 spin_unlock(&ubi->wl_lock);
681
682out_free:
683 kfree(ec_hdr);
684 return err;
685}
686
687/**
688 * check_protection_over - check if it is time to stop protecting some
689 * physical eraseblocks.
690 * @ubi: UBI device description object
691 *
692 * This function is called after each erase operation, when the absolute erase
693 * counter is incremented, to check if some physical eraseblock have not to be
694 * protected any longer. These physical eraseblocks are moved from the
695 * protection trees to the used tree.
696 */
697static void check_protection_over(struct ubi_device *ubi)
698{
699 struct ubi_wl_prot_entry *pe;
700
701 /*
702 * There may be several protected physical eraseblock to remove,
703 * process them all.
704 */
705 while (1) {
706 spin_lock(&ubi->wl_lock);
707 if (tree_empty(&ubi->prot.aec)) {
708 spin_unlock(&ubi->wl_lock);
709 break;
710 }
711
712 pe = rb_entry(rb_first(&ubi->prot.aec),
713 struct ubi_wl_prot_entry, rb_aec);
714
715 if (pe->abs_ec > ubi->abs_ec) {
716 spin_unlock(&ubi->wl_lock);
717 break;
718 }
719
720 dbg_wl("PEB %d protection over, abs_ec %llu, PEB abs_ec %llu",
721 pe->e->pnum, ubi->abs_ec, pe->abs_ec);
722 rb_erase(&pe->rb_aec, &ubi->prot.aec);
723 rb_erase(&pe->rb_pnum, &ubi->prot.pnum);
724 used_tree_add(ubi, pe->e);
725 spin_unlock(&ubi->wl_lock);
726
727 kfree(pe);
728 cond_resched();
729 }
730}
731
732/**
733 * schedule_ubi_work - schedule a work.
734 * @ubi: UBI device description object
735 * @wrk: the work to schedule
736 *
737 * This function enqueues a work defined by @wrk to the tail of the pending
738 * works list.
739 */
740static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
741{
742 spin_lock(&ubi->wl_lock);
743 list_add_tail(&wrk->list, &ubi->works);
744 ubi_assert(ubi->works_count >= 0);
745 ubi->works_count += 1;
746 if (ubi->thread_enabled)
747 wake_up_process(ubi->bgt_thread);
748 spin_unlock(&ubi->wl_lock);
749}
750
751static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
752 int cancel);
753
754/**
755 * schedule_erase - schedule an erase work.
756 * @ubi: UBI device description object
757 * @e: the WL entry of the physical eraseblock to erase
758 * @torture: if the physical eraseblock has to be tortured
759 *
760 * This function returns zero in case of success and a %-ENOMEM in case of
761 * failure.
762 */
763static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
764 int torture)
765{
766 struct ubi_work *wl_wrk;
767
768 dbg_wl("schedule erasure of PEB %d, EC %d, torture %d",
769 e->pnum, e->ec, torture);
770
771 wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_KERNEL);
772 if (!wl_wrk)
773 return -ENOMEM;
774
775 wl_wrk->func = &erase_worker;
776 wl_wrk->e = e;
777 wl_wrk->torture = torture;
778
779 schedule_ubi_work(ubi, wl_wrk);
780 return 0;
781}
782
783/**
784 * wear_leveling_worker - wear-leveling worker function.
785 * @ubi: UBI device description object
786 * @wrk: the work object
787 * @cancel: non-zero if the worker has to free memory and exit
788 *
789 * This function copies a more worn out physical eraseblock to a less worn out
790 * one. Returns zero in case of success and a negative error code in case of
791 * failure.
792 */
793static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
794 int cancel)
795{
796 int err, put = 0;
797 struct ubi_wl_entry *e1, *e2;
798 struct ubi_vid_hdr *vid_hdr;
799
800 kfree(wrk);
801
802 if (cancel)
803 return 0;
804
805 vid_hdr = ubi_zalloc_vid_hdr(ubi);
806 if (!vid_hdr)
807 return -ENOMEM;
808
809 spin_lock(&ubi->wl_lock);
810
811 /*
812 * Only one WL worker at a time is supported at this implementation, so
813 * make sure a PEB is not being moved already.
814 */
815 if (ubi->move_to || tree_empty(&ubi->free) ||
816 (tree_empty(&ubi->used) && tree_empty(&ubi->scrub))) {
817 /*
818 * Only one WL worker at a time is supported at this
819 * implementation, so if a LEB is already being moved, cancel.
820 *
821 * No free physical eraseblocks? Well, we cancel wear-leveling
822 * then. It will be triggered again when a free physical
823 * eraseblock appears.
824 *
825 * No used physical eraseblocks? They must be temporarily
826 * protected from being moved. They will be moved to the
827 * @ubi->used tree later and the wear-leveling will be
828 * triggered again.
829 */
830 dbg_wl("cancel WL, a list is empty: free %d, used %d",
831 tree_empty(&ubi->free), tree_empty(&ubi->used));
832 ubi->wl_scheduled = 0;
833 spin_unlock(&ubi->wl_lock);
834 ubi_free_vid_hdr(ubi, vid_hdr);
835 return 0;
836 }
837
838 if (tree_empty(&ubi->scrub)) {
839 /*
840 * Now pick the least worn-out used physical eraseblock and a
841 * highly worn-out free physical eraseblock. If the erase
842 * counters differ much enough, start wear-leveling.
843 */
844 e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb);
845 e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
846
847 if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) {
848 dbg_wl("no WL needed: min used EC %d, max free EC %d",
849 e1->ec, e2->ec);
850 ubi->wl_scheduled = 0;
851 spin_unlock(&ubi->wl_lock);
852 ubi_free_vid_hdr(ubi, vid_hdr);
853 return 0;
854 }
855 used_tree_del(ubi, e1);
856 dbg_wl("move PEB %d EC %d to PEB %d EC %d",
857 e1->pnum, e1->ec, e2->pnum, e2->ec);
858 } else {
859 e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, rb);
860 e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
861 scrub_tree_del(ubi, e1);
862 dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum);
863 }
864
865 free_tree_del(ubi, e2);
866 ubi_assert(!ubi->move_from && !ubi->move_to);
867 ubi_assert(!ubi->move_to_put && !ubi->move_from_put);
868 ubi->move_from = e1;
869 ubi->move_to = e2;
870 spin_unlock(&ubi->wl_lock);
871
872 /*
873 * Now we are going to copy physical eraseblock @e1->pnum to @e2->pnum.
874 * We so far do not know which logical eraseblock our physical
875 * eraseblock (@e1) belongs to. We have to read the volume identifier
876 * header first.
877 */
878
879 err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0);
880 if (err && err != UBI_IO_BITFLIPS) {
881 if (err == UBI_IO_PEB_FREE) {
882 /*
883 * We are trying to move PEB without a VID header. UBI
884 * always write VID headers shortly after the PEB was
885 * given, so we have a situation when it did not have
886 * chance to write it down because it was preempted.
887 * Just re-schedule the work, so that next time it will
888 * likely have the VID header in place.
889 */
890 dbg_wl("PEB %d has no VID header", e1->pnum);
891 err = 0;
892 } else {
893 ubi_err("error %d while reading VID header from PEB %d",
894 err, e1->pnum);
895 if (err > 0)
896 err = -EIO;
897 }
898 goto error;
899 }
900
901 err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr);
902 if (err) {
903 if (err == UBI_IO_BITFLIPS)
904 err = 0;
905 goto error;
906 }
907
908 ubi_free_vid_hdr(ubi, vid_hdr);
909 spin_lock(&ubi->wl_lock);
910 if (!ubi->move_to_put)
911 used_tree_add(ubi, e2);
912 else
913 put = 1;
914 ubi->move_from = ubi->move_to = NULL;
915 ubi->move_from_put = ubi->move_to_put = 0;
916 ubi->wl_scheduled = 0;
917 spin_unlock(&ubi->wl_lock);
918
919 if (put) {
920 /*
921 * Well, the target PEB was put meanwhile, schedule it for
922 * erasure.
923 */
924 dbg_wl("PEB %d was put meanwhile, erase", e2->pnum);
925 err = schedule_erase(ubi, e2, 0);
926 if (err) {
927 kmem_cache_free(wl_entries_slab, e2);
928 ubi_ro_mode(ubi);
929 }
930 }
931
932 err = schedule_erase(ubi, e1, 0);
933 if (err) {
934 kmem_cache_free(wl_entries_slab, e1);
935 ubi_ro_mode(ubi);
936 }
937
938 dbg_wl("done");
939 return err;
940
941 /*
942 * Some error occurred. @e1 was not changed, so return it back. @e2
943 * might be changed, schedule it for erasure.
944 */
945error:
946 if (err)
947 dbg_wl("error %d occurred, cancel operation", err);
948 ubi_assert(err <= 0);
949
950 ubi_free_vid_hdr(ubi, vid_hdr);
951 spin_lock(&ubi->wl_lock);
952 ubi->wl_scheduled = 0;
953 if (ubi->move_from_put)
954 put = 1;
955 else
956 used_tree_add(ubi, e1);
957 ubi->move_from = ubi->move_to = NULL;
958 ubi->move_from_put = ubi->move_to_put = 0;
959 spin_unlock(&ubi->wl_lock);
960
961 if (put) {
962 /*
963 * Well, the target PEB was put meanwhile, schedule it for
964 * erasure.
965 */
966 dbg_wl("PEB %d was put meanwhile, erase", e1->pnum);
967 err = schedule_erase(ubi, e1, 0);
968 if (err) {
969 kmem_cache_free(wl_entries_slab, e1);
970 ubi_ro_mode(ubi);
971 }
972 }
973
974 err = schedule_erase(ubi, e2, 0);
975 if (err) {
976 kmem_cache_free(wl_entries_slab, e2);
977 ubi_ro_mode(ubi);
978 }
979
980 yield();
981 return err;
982}
983
984/**
985 * ensure_wear_leveling - schedule wear-leveling if it is needed.
986 * @ubi: UBI device description object
987 *
988 * This function checks if it is time to start wear-leveling and schedules it
989 * if yes. This function returns zero in case of success and a negative error
990 * code in case of failure.
991 */
992static int ensure_wear_leveling(struct ubi_device *ubi)
993{
994 int err = 0;
995 struct ubi_wl_entry *e1;
996 struct ubi_wl_entry *e2;
997 struct ubi_work *wrk;
998
999 spin_lock(&ubi->wl_lock);
1000 if (ubi->wl_scheduled)
1001 /* Wear-leveling is already in the work queue */
1002 goto out_unlock;
1003
1004 /*
1005 * If the ubi->scrub tree is not empty, scrubbing is needed, and the
1006 * the WL worker has to be scheduled anyway.
1007 */
1008 if (tree_empty(&ubi->scrub)) {
1009 if (tree_empty(&ubi->used) || tree_empty(&ubi->free))
1010 /* No physical eraseblocks - no deal */
1011 goto out_unlock;
1012
1013 /*
1014 * We schedule wear-leveling only if the difference between the
1015 * lowest erase counter of used physical eraseblocks and a high
1016 * erase counter of free physical eraseblocks is greater then
1017 * %UBI_WL_THRESHOLD.
1018 */
1019 e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb);
1020 e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
1021
1022 if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD))
1023 goto out_unlock;
1024 dbg_wl("schedule wear-leveling");
1025 } else
1026 dbg_wl("schedule scrubbing");
1027
1028 ubi->wl_scheduled = 1;
1029 spin_unlock(&ubi->wl_lock);
1030
1031 wrk = kmalloc(sizeof(struct ubi_work), GFP_KERNEL);
1032 if (!wrk) {
1033 err = -ENOMEM;
1034 goto out_cancel;
1035 }
1036
1037 wrk->func = &wear_leveling_worker;
1038 schedule_ubi_work(ubi, wrk);
1039 return err;
1040
1041out_cancel:
1042 spin_lock(&ubi->wl_lock);
1043 ubi->wl_scheduled = 0;
1044out_unlock:
1045 spin_unlock(&ubi->wl_lock);
1046 return err;
1047}
1048
1049/**
1050 * erase_worker - physical eraseblock erase worker function.
1051 * @ubi: UBI device description object
1052 * @wl_wrk: the work object
1053 * @cancel: non-zero if the worker has to free memory and exit
1054 *
1055 * This function erases a physical eraseblock and perform torture testing if
1056 * needed. It also takes care about marking the physical eraseblock bad if
1057 * needed. Returns zero in case of success and a negative error code in case of
1058 * failure.
1059 */
1060static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
1061 int cancel)
1062{
1063 int err;
1064 struct ubi_wl_entry *e = wl_wrk->e;
1065 int pnum = e->pnum;
1066
1067 if (cancel) {
1068 dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec);
1069 kfree(wl_wrk);
1070 kmem_cache_free(wl_entries_slab, e);
1071 return 0;
1072 }
1073
1074 dbg_wl("erase PEB %d EC %d", pnum, e->ec);
1075
1076 err = sync_erase(ubi, e, wl_wrk->torture);
1077 if (!err) {
1078 /* Fine, we've erased it successfully */
1079 kfree(wl_wrk);
1080
1081 spin_lock(&ubi->wl_lock);
1082 ubi->abs_ec += 1;
1083 free_tree_add(ubi, e);
1084 spin_unlock(&ubi->wl_lock);
1085
1086 /*
1087 * One more erase operation has happened, take care about protected
1088 * physical eraseblocks.
1089 */
1090 check_protection_over(ubi);
1091
1092 /* And take care about wear-leveling */
1093 err = ensure_wear_leveling(ubi);
1094 return err;
1095 }
1096
1097 kfree(wl_wrk);
1098 kmem_cache_free(wl_entries_slab, e);
1099
1100 if (err != -EIO) {
1101 /*
1102 * If this is not %-EIO, we have no idea what to do. Scheduling
1103 * this physical eraseblock for erasure again would cause
1104 * errors again and again. Well, lets switch to RO mode.
1105 */
1106 ubi_ro_mode(ubi);
1107 return err;
1108 }
1109
1110 /* It is %-EIO, the PEB went bad */
1111
1112 if (!ubi->bad_allowed) {
1113 ubi_err("bad physical eraseblock %d detected", pnum);
1114 ubi_ro_mode(ubi);
1115 err = -EIO;
1116 } else {
1117 int need;
1118
1119 spin_lock(&ubi->volumes_lock);
1120 need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1;
1121 if (need > 0) {
1122 need = ubi->avail_pebs >= need ? need : ubi->avail_pebs;
1123 ubi->avail_pebs -= need;
1124 ubi->rsvd_pebs += need;
1125 ubi->beb_rsvd_pebs += need;
1126 if (need > 0)
1127 ubi_msg("reserve more %d PEBs", need);
1128 }
1129
1130 if (ubi->beb_rsvd_pebs == 0) {
1131 spin_unlock(&ubi->volumes_lock);
1132 ubi_err("no reserved physical eraseblocks");
1133 ubi_ro_mode(ubi);
1134 return -EIO;
1135 }
1136
1137 spin_unlock(&ubi->volumes_lock);
1138 ubi_msg("mark PEB %d as bad", pnum);
1139
1140 err = ubi_io_mark_bad(ubi, pnum);
1141 if (err) {
1142 ubi_ro_mode(ubi);
1143 return err;
1144 }
1145
1146 spin_lock(&ubi->volumes_lock);
1147 ubi->beb_rsvd_pebs -= 1;
1148 ubi->bad_peb_count += 1;
1149 ubi->good_peb_count -= 1;
1150 ubi_calculate_reserved(ubi);
1151 if (ubi->beb_rsvd_pebs == 0)
1152 ubi_warn("last PEB from the reserved pool was used");
1153 spin_unlock(&ubi->volumes_lock);
1154 }
1155
1156 return err;
1157}
1158
1159/**
1160 * ubi_wl_put_peb - return a physical eraseblock to the wear-leveling
1161 * unit.
1162 * @ubi: UBI device description object
1163 * @pnum: physical eraseblock to return
1164 * @torture: if this physical eraseblock has to be tortured
1165 *
1166 * This function is called to return physical eraseblock @pnum to the pool of
1167 * free physical eraseblocks. The @torture flag has to be set if an I/O error
1168 * occurred to this @pnum and it has to be tested. This function returns zero
1169 * in case of success and a negative error code in case of failure.
1170 */
1171int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture)
1172{
1173 int err;
1174 struct ubi_wl_entry *e;
1175
1176 dbg_wl("PEB %d", pnum);
1177 ubi_assert(pnum >= 0);
1178 ubi_assert(pnum < ubi->peb_count);
1179
1180 spin_lock(&ubi->wl_lock);
1181
1182 e = ubi->lookuptbl[pnum];
1183 if (e == ubi->move_from) {
1184 /*
1185 * User is putting the physical eraseblock which was selected to
1186 * be moved. It will be scheduled for erasure in the
1187 * wear-leveling worker.
1188 */
1189 dbg_wl("PEB %d is being moved", pnum);
1190 ubi_assert(!ubi->move_from_put);
1191 ubi->move_from_put = 1;
1192 spin_unlock(&ubi->wl_lock);
1193 return 0;
1194 } else if (e == ubi->move_to) {
1195 /*
1196 * User is putting the physical eraseblock which was selected
1197 * as the target the data is moved to. It may happen if the EBA
1198 * unit already re-mapped the LEB but the WL unit did has not
1199 * put the PEB to the "used" tree.
1200 */
1201 dbg_wl("PEB %d is the target of data moving", pnum);
1202 ubi_assert(!ubi->move_to_put);
1203 ubi->move_to_put = 1;
1204 spin_unlock(&ubi->wl_lock);
1205 return 0;
1206 } else {
1207 if (in_wl_tree(e, &ubi->used))
1208 used_tree_del(ubi, e);
1209 else if (in_wl_tree(e, &ubi->scrub))
1210 scrub_tree_del(ubi, e);
1211 else
1212 prot_tree_del(ubi, e->pnum);
1213 }
1214 spin_unlock(&ubi->wl_lock);
1215
1216 err = schedule_erase(ubi, e, torture);
1217 if (err) {
1218 spin_lock(&ubi->wl_lock);
1219 used_tree_add(ubi, e);
1220 spin_unlock(&ubi->wl_lock);
1221 }
1222
1223 return err;
1224}
1225
1226/**
1227 * ubi_wl_scrub_peb - schedule a physical eraseblock for scrubbing.
1228 * @ubi: UBI device description object
1229 * @pnum: the physical eraseblock to schedule
1230 *
1231 * If a bit-flip in a physical eraseblock is detected, this physical eraseblock
1232 * needs scrubbing. This function schedules a physical eraseblock for
1233 * scrubbing which is done in background. This function returns zero in case of
1234 * success and a negative error code in case of failure.
1235 */
1236int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum)
1237{
1238 struct ubi_wl_entry *e;
1239
1240 ubi_msg("schedule PEB %d for scrubbing", pnum);
1241
1242retry:
1243 spin_lock(&ubi->wl_lock);
1244 e = ubi->lookuptbl[pnum];
1245 if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub)) {
1246 spin_unlock(&ubi->wl_lock);
1247 return 0;
1248 }
1249
1250 if (e == ubi->move_to) {
1251 /*
1252 * This physical eraseblock was used to move data to. The data
1253 * was moved but the PEB was not yet inserted to the proper
1254 * tree. We should just wait a little and let the WL worker
1255 * proceed.
1256 */
1257 spin_unlock(&ubi->wl_lock);
1258 dbg_wl("the PEB %d is not in proper tree, retry", pnum);
1259 yield();
1260 goto retry;
1261 }
1262
1263 if (in_wl_tree(e, &ubi->used))
1264 used_tree_del(ubi, e);
1265 else
1266 prot_tree_del(ubi, pnum);
1267
1268 scrub_tree_add(ubi, e);
1269 spin_unlock(&ubi->wl_lock);
1270
1271 /*
1272 * Technically scrubbing is the same as wear-leveling, so it is done
1273 * by the WL worker.
1274 */
1275 return ensure_wear_leveling(ubi);
1276}
1277
1278/**
1279 * ubi_wl_flush - flush all pending works.
1280 * @ubi: UBI device description object
1281 *
1282 * This function returns zero in case of success and a negative error code in
1283 * case of failure.
1284 */
1285int ubi_wl_flush(struct ubi_device *ubi)
1286{
1287 int err, pending_count;
1288
1289 pending_count = ubi->works_count;
1290
1291 dbg_wl("flush (%d pending works)", pending_count);
1292
1293 /*
1294 * Erase while the pending works queue is not empty, but not more then
1295 * the number of currently pending works.
1296 */
1297 while (pending_count-- > 0) {
1298 err = do_work(ubi);
1299 if (err)
1300 return err;
1301 }
1302
1303 return 0;
1304}
1305
1306/**
1307 * tree_destroy - destroy an RB-tree.
1308 * @root: the root of the tree to destroy
1309 */
1310static void tree_destroy(struct rb_root *root)
1311{
1312 struct rb_node *rb;
1313 struct ubi_wl_entry *e;
1314
1315 rb = root->rb_node;
1316 while (rb) {
1317 if (rb->rb_left)
1318 rb = rb->rb_left;
1319 else if (rb->rb_right)
1320 rb = rb->rb_right;
1321 else {
1322 e = rb_entry(rb, struct ubi_wl_entry, rb);
1323
1324 rb = rb_parent(rb);
1325 if (rb) {
1326 if (rb->rb_left == &e->rb)
1327 rb->rb_left = NULL;
1328 else
1329 rb->rb_right = NULL;
1330 }
1331
1332 kmem_cache_free(wl_entries_slab, e);
1333 }
1334 }
1335}
1336
1337/**
1338 * ubi_thread - UBI background thread.
1339 * @u: the UBI device description object pointer
1340 */
1341static int ubi_thread(void *u)
1342{
1343 int failures = 0;
1344 struct ubi_device *ubi = u;
1345
1346 ubi_msg("background thread \"%s\" started, PID %d",
1347 ubi->bgt_name, current->pid);
1348
1349 for (;;) {
1350 int err;
1351
1352 if (kthread_should_stop())
1353 goto out;
1354
1355 if (try_to_freeze())
1356 continue;
1357
1358 spin_lock(&ubi->wl_lock);
1359 if (list_empty(&ubi->works) || ubi->ro_mode ||
1360 !ubi->thread_enabled) {
1361 set_current_state(TASK_INTERRUPTIBLE);
1362 spin_unlock(&ubi->wl_lock);
1363 schedule();
1364 continue;
1365 }
1366 spin_unlock(&ubi->wl_lock);
1367
1368 err = do_work(ubi);
1369 if (err) {
1370 ubi_err("%s: work failed with error code %d",
1371 ubi->bgt_name, err);
1372 if (failures++ > WL_MAX_FAILURES) {
1373 /*
1374 * Too many failures, disable the thread and
1375 * switch to read-only mode.
1376 */
1377 ubi_msg("%s: %d consecutive failures",
1378 ubi->bgt_name, WL_MAX_FAILURES);
1379 ubi_ro_mode(ubi);
1380 break;
1381 }
1382 } else
1383 failures = 0;
1384
1385 cond_resched();
1386 }
1387
1388out:
1389 dbg_wl("background thread \"%s\" is killed", ubi->bgt_name);
1390 return 0;
1391}
1392
1393/**
1394 * cancel_pending - cancel all pending works.
1395 * @ubi: UBI device description object
1396 */
1397static void cancel_pending(struct ubi_device *ubi)
1398{
1399 while (!list_empty(&ubi->works)) {
1400 struct ubi_work *wrk;
1401
1402 wrk = list_entry(ubi->works.next, struct ubi_work, list);
1403 list_del(&wrk->list);
1404 wrk->func(ubi, wrk, 1);
1405 ubi->works_count -= 1;
1406 ubi_assert(ubi->works_count >= 0);
1407 }
1408}
1409
1410/**
1411 * ubi_wl_init_scan - initialize the wear-leveling unit using scanning
1412 * information.
1413 * @ubi: UBI device description object
1414 * @si: scanning information
1415 *
1416 * This function returns zero in case of success, and a negative error code in
1417 * case of failure.
1418 */
1419int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1420{
1421 int err;
1422 struct rb_node *rb1, *rb2;
1423 struct ubi_scan_volume *sv;
1424 struct ubi_scan_leb *seb, *tmp;
1425 struct ubi_wl_entry *e;
1426
1427
1428 ubi->used = ubi->free = ubi->scrub = RB_ROOT;
1429 ubi->prot.pnum = ubi->prot.aec = RB_ROOT;
1430 spin_lock_init(&ubi->wl_lock);
1431 ubi->max_ec = si->max_ec;
1432 INIT_LIST_HEAD(&ubi->works);
1433
1434 sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num);
1435
1436 ubi->bgt_thread = kthread_create(ubi_thread, ubi, ubi->bgt_name);
1437 if (IS_ERR(ubi->bgt_thread)) {
1438 err = PTR_ERR(ubi->bgt_thread);
1439 ubi_err("cannot spawn \"%s\", error %d", ubi->bgt_name,
1440 err);
1441 return err;
1442 }
1443
1444 if (ubi_devices_cnt == 0) {
1445 wl_entries_slab = kmem_cache_create("ubi_wl_entry_slab",
1446 sizeof(struct ubi_wl_entry),
1447 0, 0, NULL, NULL);
1448 if (!wl_entries_slab)
1449 return -ENOMEM;
1450 }
1451
1452 err = -ENOMEM;
1453 ubi->lookuptbl = kzalloc(ubi->peb_count * sizeof(void *), GFP_KERNEL);
1454 if (!ubi->lookuptbl)
1455 goto out_free;
1456
1457 list_for_each_entry_safe(seb, tmp, &si->erase, u.list) {
1458 cond_resched();
1459
1460 e = kmem_cache_alloc(wl_entries_slab, GFP_KERNEL);
1461 if (!e)
1462 goto out_free;
1463
1464 e->pnum = seb->pnum;
1465 e->ec = seb->ec;
1466 ubi->lookuptbl[e->pnum] = e;
1467 if (schedule_erase(ubi, e, 0)) {
1468 kmem_cache_free(wl_entries_slab, e);
1469 goto out_free;
1470 }
1471 }
1472
1473 list_for_each_entry(seb, &si->free, u.list) {
1474 cond_resched();
1475
1476 e = kmem_cache_alloc(wl_entries_slab, GFP_KERNEL);
1477 if (!e)
1478 goto out_free;
1479
1480 e->pnum = seb->pnum;
1481 e->ec = seb->ec;
1482 ubi_assert(e->ec >= 0);
1483 free_tree_add(ubi, e);
1484 ubi->lookuptbl[e->pnum] = e;
1485 }
1486
1487 list_for_each_entry(seb, &si->corr, u.list) {
1488 cond_resched();
1489
1490 e = kmem_cache_alloc(wl_entries_slab, GFP_KERNEL);
1491 if (!e)
1492 goto out_free;
1493
1494 e->pnum = seb->pnum;
1495 e->ec = seb->ec;
1496 ubi->lookuptbl[e->pnum] = e;
1497 if (schedule_erase(ubi, e, 0)) {
1498 kmem_cache_free(wl_entries_slab, e);
1499 goto out_free;
1500 }
1501 }
1502
1503 ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
1504 ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) {
1505 cond_resched();
1506
1507 e = kmem_cache_alloc(wl_entries_slab, GFP_KERNEL);
1508 if (!e)
1509 goto out_free;
1510
1511 e->pnum = seb->pnum;
1512 e->ec = seb->ec;
1513 ubi->lookuptbl[e->pnum] = e;
1514 if (!seb->scrub) {
1515 dbg_wl("add PEB %d EC %d to the used tree",
1516 e->pnum, e->ec);
1517 used_tree_add(ubi, e);
1518 } else {
1519 dbg_wl("add PEB %d EC %d to the scrub tree",
1520 e->pnum, e->ec);
1521 scrub_tree_add(ubi, e);
1522 }
1523 }
1524 }
1525
1526 if (WL_RESERVED_PEBS > ubi->avail_pebs) {
1527 ubi_err("no enough physical eraseblocks (%d, need %d)",
1528 ubi->avail_pebs, WL_RESERVED_PEBS);
1529 goto out_free;
1530 }
1531 ubi->avail_pebs -= WL_RESERVED_PEBS;
1532 ubi->rsvd_pebs += WL_RESERVED_PEBS;
1533
1534 /* Schedule wear-leveling if needed */
1535 err = ensure_wear_leveling(ubi);
1536 if (err)
1537 goto out_free;
1538
1539 return 0;
1540
1541out_free:
1542 cancel_pending(ubi);
1543 tree_destroy(&ubi->used);
1544 tree_destroy(&ubi->free);
1545 tree_destroy(&ubi->scrub);
1546 kfree(ubi->lookuptbl);
1547 if (ubi_devices_cnt == 0)
1548 kmem_cache_destroy(wl_entries_slab);
1549 return err;
1550}
1551
1552/**
1553 * protection_trees_destroy - destroy the protection RB-trees.
1554 * @ubi: UBI device description object
1555 */
1556static void protection_trees_destroy(struct ubi_device *ubi)
1557{
1558 struct rb_node *rb;
1559 struct ubi_wl_prot_entry *pe;
1560
1561 rb = ubi->prot.aec.rb_node;
1562 while (rb) {
1563 if (rb->rb_left)
1564 rb = rb->rb_left;
1565 else if (rb->rb_right)
1566 rb = rb->rb_right;
1567 else {
1568 pe = rb_entry(rb, struct ubi_wl_prot_entry, rb_aec);
1569
1570 rb = rb_parent(rb);
1571 if (rb) {
1572 if (rb->rb_left == &pe->rb_aec)
1573 rb->rb_left = NULL;
1574 else
1575 rb->rb_right = NULL;
1576 }
1577
1578 kmem_cache_free(wl_entries_slab, pe->e);
1579 kfree(pe);
1580 }
1581 }
1582}
1583
1584/**
1585 * ubi_wl_close - close the wear-leveling unit.
1586 * @ubi: UBI device description object
1587 */
1588void ubi_wl_close(struct ubi_device *ubi)
1589{
1590 dbg_wl("disable \"%s\"", ubi->bgt_name);
1591 if (ubi->bgt_thread)
1592 kthread_stop(ubi->bgt_thread);
1593
1594 dbg_wl("close the UBI wear-leveling unit");
1595
1596 cancel_pending(ubi);
1597 protection_trees_destroy(ubi);
1598 tree_destroy(&ubi->used);
1599 tree_destroy(&ubi->free);
1600 tree_destroy(&ubi->scrub);
1601 kfree(ubi->lookuptbl);
1602 if (ubi_devices_cnt == 1)
1603 kmem_cache_destroy(wl_entries_slab);
1604}
1605
1606#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
1607
1608/**
1609 * paranoid_check_ec - make sure that the erase counter of a physical eraseblock
1610 * is correct.
1611 * @ubi: UBI device description object
1612 * @pnum: the physical eraseblock number to check
1613 * @ec: the erase counter to check
1614 *
1615 * This function returns zero if the erase counter of physical eraseblock @pnum
1616 * is equivalent to @ec, %1 if not, and a negative error code if an error
1617 * occurred.
1618 */
1619static int paranoid_check_ec(const struct ubi_device *ubi, int pnum, int ec)
1620{
1621 int err;
1622 long long read_ec;
1623 struct ubi_ec_hdr *ec_hdr;
1624
1625 ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
1626 if (!ec_hdr)
1627 return -ENOMEM;
1628
1629 err = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0);
1630 if (err && err != UBI_IO_BITFLIPS) {
1631 /* The header does not have to exist */
1632 err = 0;
1633 goto out_free;
1634 }
1635
1636 read_ec = ubi64_to_cpu(ec_hdr->ec);
1637 if (ec != read_ec) {
1638 ubi_err("paranoid check failed for PEB %d", pnum);
1639 ubi_err("read EC is %lld, should be %d", read_ec, ec);
1640 ubi_dbg_dump_stack();
1641 err = 1;
1642 } else
1643 err = 0;
1644
1645out_free:
1646 kfree(ec_hdr);
1647 return err;
1648}
1649
1650/**
1651 * paranoid_check_in_wl_tree - make sure that a wear-leveling entry is present
1652 * in a WL RB-tree.
1653 * @e: the wear-leveling entry to check
1654 * @root: the root of the tree
1655 *
1656 * This function returns zero if @e is in the @root RB-tree and %1 if it
1657 * is not.
1658 */
1659static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e,
1660 struct rb_root *root)
1661{
1662 if (in_wl_tree(e, root))
1663 return 0;
1664
1665 ubi_err("paranoid check failed for PEB %d, EC %d, RB-tree %p ",
1666 e->pnum, e->ec, root);
1667 ubi_dbg_dump_stack();
1668 return 1;
1669}
1670
1671#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
diff --git a/include/linux/mtd/ubi.h b/include/linux/mtd/ubi.h
new file mode 100644
index 000000000000..3d967b6b120a
--- /dev/null
+++ b/include/linux/mtd/ubi.h
@@ -0,0 +1,202 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21#ifndef __LINUX_UBI_H__
22#define __LINUX_UBI_H__
23
24#include <asm/ioctl.h>
25#include <linux/types.h>
26#include <mtd/ubi-user.h>
27
28/*
29 * UBI data type hint constants.
30 *
31 * UBI_LONGTERM: long-term data
32 * UBI_SHORTTERM: short-term data
33 * UBI_UNKNOWN: data persistence is unknown
34 *
35 * These constants are used when data is written to UBI volumes in order to
36 * help the UBI wear-leveling unit to find more appropriate physical
37 * eraseblocks.
38 */
39enum {
40 UBI_LONGTERM = 1,
41 UBI_SHORTTERM,
42 UBI_UNKNOWN
43};
44
45/*
46 * enum ubi_open_mode - UBI volume open mode constants.
47 *
48 * UBI_READONLY: read-only mode
49 * UBI_READWRITE: read-write mode
50 * UBI_EXCLUSIVE: exclusive mode
51 */
52enum {
53 UBI_READONLY = 1,
54 UBI_READWRITE,
55 UBI_EXCLUSIVE
56};
57
58/**
59 * struct ubi_volume_info - UBI volume description data structure.
60 * @vol_id: volume ID
61 * @ubi_num: UBI device number this volume belongs to
62 * @size: how many physical eraseblocks are reserved for this volume
63 * @used_bytes: how many bytes of data this volume contains
64 * @used_ebs: how many physical eraseblocks of this volume actually contain any
65 * data
66 * @vol_type: volume type (%UBI_DYNAMIC_VOLUME or %UBI_STATIC_VOLUME)
67 * @corrupted: non-zero if the volume is corrupted (static volumes only)
68 * @upd_marker: non-zero if the volume has update marker set
69 * @alignment: volume alignment
70 * @usable_leb_size: how many bytes are available in logical eraseblocks of
71 * this volume
72 * @name_len: volume name length
73 * @name: volume name
74 * @cdev: UBI volume character device major and minor numbers
75 *
76 * The @corrupted flag is only relevant to static volumes and is always zero
77 * for dynamic ones. This is because UBI does not care about dynamic volume
78 * data protection and only cares about protecting static volume data.
79 *
80 * The @upd_marker flag is set if the volume update operation was interrupted.
81 * Before touching the volume data during the update operation, UBI first sets
82 * the update marker flag for this volume. If the volume update operation was
83 * further interrupted, the update marker indicates this. If the update marker
84 * is set, the contents of the volume is certainly damaged and a new volume
85 * update operation has to be started.
86 *
87 * To put it differently, @corrupted and @upd_marker fields have different
88 * semantics:
89 * o the @corrupted flag means that this static volume is corrupted for some
90 * reasons, but not because an interrupted volume update
91 * o the @upd_marker field means that the volume is damaged because of an
92 * interrupted update operation.
93 *
94 * I.e., the @corrupted flag is never set if the @upd_marker flag is set.
95 *
96 * The @used_bytes and @used_ebs fields are only really needed for static
97 * volumes and contain the number of bytes stored in this static volume and how
98 * many eraseblock this data occupies. In case of dynamic volumes, the
99 * @used_bytes field is equivalent to @size*@usable_leb_size, and the @used_ebs
100 * field is equivalent to @size.
101 *
102 * In general, logical eraseblock size is a property of the UBI device, not
103 * of the UBI volume. Indeed, the logical eraseblock size depends on the
104 * physical eraseblock size and on how much bytes UBI headers consume. But
105 * because of the volume alignment (@alignment), the usable size of logical
106 * eraseblocks if a volume may be less. The following equation is true:
107 * @usable_leb_size = LEB size - (LEB size mod @alignment),
108 * where LEB size is the logical eraseblock size defined by the UBI device.
109 *
110 * The alignment is multiple to the minimal flash input/output unit size or %1
111 * if all the available space is used.
112 *
113 * To put this differently, alignment may be considered is a way to change
114 * volume logical eraseblock sizes.
115 */
116struct ubi_volume_info {
117 int ubi_num;
118 int vol_id;
119 int size;
120 long long used_bytes;
121 int used_ebs;
122 int vol_type;
123 int corrupted;
124 int upd_marker;
125 int alignment;
126 int usable_leb_size;
127 int name_len;
128 const char *name;
129 dev_t cdev;
130};
131
132/**
133 * struct ubi_device_info - UBI device description data structure.
134 * @ubi_num: ubi device number
135 * @leb_size: logical eraseblock size on this UBI device
136 * @min_io_size: minimal I/O unit size
137 * @ro_mode: if this device is in read-only mode
138 * @cdev: UBI character device major and minor numbers
139 *
140 * Note, @leb_size is the logical eraseblock size offered by the UBI device.
141 * Volumes of this UBI device may have smaller logical eraseblock size if their
142 * alignment is not equivalent to %1.
143 */
144struct ubi_device_info {
145 int ubi_num;
146 int leb_size;
147 int min_io_size;
148 int ro_mode;
149 dev_t cdev;
150};
151
152/* UBI descriptor given to users when they open UBI volumes */
153struct ubi_volume_desc;
154
155int ubi_get_device_info(int ubi_num, struct ubi_device_info *di);
156void ubi_get_volume_info(struct ubi_volume_desc *desc,
157 struct ubi_volume_info *vi);
158struct ubi_volume_desc *ubi_open_volume(int ubi_num, int vol_id, int mode);
159struct ubi_volume_desc *ubi_open_volume_nm(int ubi_num, const char *name,
160 int mode);
161void ubi_close_volume(struct ubi_volume_desc *desc);
162int ubi_leb_read(struct ubi_volume_desc *desc, int lnum, char *buf, int offset,
163 int len, int check);
164int ubi_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf,
165 int offset, int len, int dtype);
166int ubi_leb_change(struct ubi_volume_desc *desc, int lnum, const void *buf,
167 int len, int dtype);
168int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum);
169int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum);
170int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum);
171
172/*
173 * This function is the same as the 'ubi_leb_read()' function, but it does not
174 * provide the checking capability.
175 */
176static inline int ubi_read(struct ubi_volume_desc *desc, int lnum, char *buf,
177 int offset, int len)
178{
179 return ubi_leb_read(desc, lnum, buf, offset, len, 0);
180}
181
182/*
183 * This function is the same as the 'ubi_leb_write()' functions, but it does
184 * not have the data type argument.
185 */
186static inline int ubi_write(struct ubi_volume_desc *desc, int lnum,
187 const void *buf, int offset, int len)
188{
189 return ubi_leb_write(desc, lnum, buf, offset, len, UBI_UNKNOWN);
190}
191
192/*
193 * This function is the same as the 'ubi_leb_change()' functions, but it does
194 * not have the data type argument.
195 */
196static inline int ubi_change(struct ubi_volume_desc *desc, int lnum,
197 const void *buf, int len)
198{
199 return ubi_leb_change(desc, lnum, buf, len, UBI_UNKNOWN);
200}
201
202#endif /* !__LINUX_UBI_H__ */
diff --git a/include/mtd/Kbuild b/include/mtd/Kbuild
index e0fe92b03a4e..4d46b3bdebd8 100644
--- a/include/mtd/Kbuild
+++ b/include/mtd/Kbuild
@@ -3,3 +3,5 @@ header-y += jffs2-user.h
3header-y += mtd-abi.h 3header-y += mtd-abi.h
4header-y += mtd-user.h 4header-y += mtd-user.h
5header-y += nftl-user.h 5header-y += nftl-user.h
6header-y += ubi-header.h
7header-y += ubi-user.h
diff --git a/include/mtd/mtd-abi.h b/include/mtd/mtd-abi.h
index 8e501a75a764..f71dac420394 100644
--- a/include/mtd/mtd-abi.h
+++ b/include/mtd/mtd-abi.h
@@ -24,6 +24,7 @@ struct mtd_oob_buf {
24#define MTD_NORFLASH 3 24#define MTD_NORFLASH 3
25#define MTD_NANDFLASH 4 25#define MTD_NANDFLASH 4
26#define MTD_DATAFLASH 6 26#define MTD_DATAFLASH 6
27#define MTD_UBIVOLUME 7
27 28
28#define MTD_WRITEABLE 0x400 /* Device is writeable */ 29#define MTD_WRITEABLE 0x400 /* Device is writeable */
29#define MTD_BIT_WRITEABLE 0x800 /* Single bits can be flipped */ 30#define MTD_BIT_WRITEABLE 0x800 /* Single bits can be flipped */
diff --git a/include/mtd/ubi-header.h b/include/mtd/ubi-header.h
new file mode 100644
index 000000000000..fa479c71aa34
--- /dev/null
+++ b/include/mtd/ubi-header.h
@@ -0,0 +1,360 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Authors: Artem Bityutskiy (Битюцкий Артём)
19 * Thomas Gleixner
20 * Frank Haverkamp
21 * Oliver Lohmann
22 * Andreas Arnez
23 */
24
25/*
26 * This file defines the layout of UBI headers and all the other UBI on-flash
27 * data structures. May be included by user-space.
28 */
29
30#ifndef __UBI_HEADER_H__
31#define __UBI_HEADER_H__
32
33#include <asm/byteorder.h>
34
35/* The version of UBI images supported by this implementation */
36#define UBI_VERSION 1
37
38/* The highest erase counter value supported by this implementation */
39#define UBI_MAX_ERASECOUNTER 0x7FFFFFFF
40
41/* The initial CRC32 value used when calculating CRC checksums */
42#define UBI_CRC32_INIT 0xFFFFFFFFU
43
44/* Erase counter header magic number (ASCII "UBI#") */
45#define UBI_EC_HDR_MAGIC 0x55424923
46/* Volume identifier header magic number (ASCII "UBI!") */
47#define UBI_VID_HDR_MAGIC 0x55424921
48
49/*
50 * Volume type constants used in the volume identifier header.
51 *
52 * @UBI_VID_DYNAMIC: dynamic volume
53 * @UBI_VID_STATIC: static volume
54 */
55enum {
56 UBI_VID_DYNAMIC = 1,
57 UBI_VID_STATIC = 2
58};
59
60/*
61 * Compatibility constants used by internal volumes.
62 *
63 * @UBI_COMPAT_DELETE: delete this internal volume before anything is written
64 * to the flash
65 * @UBI_COMPAT_RO: attach this device in read-only mode
66 * @UBI_COMPAT_PRESERVE: preserve this internal volume - do not touch its
67 * physical eraseblocks, don't allow the wear-leveling unit to move them
68 * @UBI_COMPAT_REJECT: reject this UBI image
69 */
70enum {
71 UBI_COMPAT_DELETE = 1,
72 UBI_COMPAT_RO = 2,
73 UBI_COMPAT_PRESERVE = 4,
74 UBI_COMPAT_REJECT = 5
75};
76
77/*
78 * ubi16_t/ubi32_t/ubi64_t - 16, 32, and 64-bit integers used in UBI on-flash
79 * data structures.
80 */
81typedef struct {
82 uint16_t int16;
83} __attribute__ ((packed)) ubi16_t;
84
85typedef struct {
86 uint32_t int32;
87} __attribute__ ((packed)) ubi32_t;
88
89typedef struct {
90 uint64_t int64;
91} __attribute__ ((packed)) ubi64_t;
92
93/*
94 * In this implementation of UBI uses the big-endian format for on-flash
95 * integers. The below are the corresponding conversion macros.
96 */
97#define cpu_to_ubi16(x) ((ubi16_t){__cpu_to_be16(x)})
98#define ubi16_to_cpu(x) ((uint16_t)__be16_to_cpu((x).int16))
99
100#define cpu_to_ubi32(x) ((ubi32_t){__cpu_to_be32(x)})
101#define ubi32_to_cpu(x) ((uint32_t)__be32_to_cpu((x).int32))
102
103#define cpu_to_ubi64(x) ((ubi64_t){__cpu_to_be64(x)})
104#define ubi64_to_cpu(x) ((uint64_t)__be64_to_cpu((x).int64))
105
106/* Sizes of UBI headers */
107#define UBI_EC_HDR_SIZE sizeof(struct ubi_ec_hdr)
108#define UBI_VID_HDR_SIZE sizeof(struct ubi_vid_hdr)
109
110/* Sizes of UBI headers without the ending CRC */
111#define UBI_EC_HDR_SIZE_CRC (UBI_EC_HDR_SIZE - sizeof(ubi32_t))
112#define UBI_VID_HDR_SIZE_CRC (UBI_VID_HDR_SIZE - sizeof(ubi32_t))
113
114/**
115 * struct ubi_ec_hdr - UBI erase counter header.
116 * @magic: erase counter header magic number (%UBI_EC_HDR_MAGIC)
117 * @version: version of UBI implementation which is supposed to accept this
118 * UBI image
119 * @padding1: reserved for future, zeroes
120 * @ec: the erase counter
121 * @vid_hdr_offset: where the VID header starts
122 * @data_offset: where the user data start
123 * @padding2: reserved for future, zeroes
124 * @hdr_crc: erase counter header CRC checksum
125 *
126 * The erase counter header takes 64 bytes and has a plenty of unused space for
127 * future usage. The unused fields are zeroed. The @version field is used to
128 * indicate the version of UBI implementation which is supposed to be able to
129 * work with this UBI image. If @version is greater then the current UBI
130 * version, the image is rejected. This may be useful in future if something
131 * is changed radically. This field is duplicated in the volume identifier
132 * header.
133 *
134 * The @vid_hdr_offset and @data_offset fields contain the offset of the the
135 * volume identifier header and user data, relative to the beginning of the
136 * physical eraseblock. These values have to be the same for all physical
137 * eraseblocks.
138 */
139struct ubi_ec_hdr {
140 ubi32_t magic;
141 uint8_t version;
142 uint8_t padding1[3];
143 ubi64_t ec; /* Warning: the current limit is 31-bit anyway! */
144 ubi32_t vid_hdr_offset;
145 ubi32_t data_offset;
146 uint8_t padding2[36];
147 ubi32_t hdr_crc;
148} __attribute__ ((packed));
149
150/**
151 * struct ubi_vid_hdr - on-flash UBI volume identifier header.
152 * @magic: volume identifier header magic number (%UBI_VID_HDR_MAGIC)
153 * @version: UBI implementation version which is supposed to accept this UBI
154 * image (%UBI_VERSION)
155 * @vol_type: volume type (%UBI_VID_DYNAMIC or %UBI_VID_STATIC)
156 * @copy_flag: if this logical eraseblock was copied from another physical
157 * eraseblock (for wear-leveling reasons)
158 * @compat: compatibility of this volume (%0, %UBI_COMPAT_DELETE,
159 * %UBI_COMPAT_IGNORE, %UBI_COMPAT_PRESERVE, or %UBI_COMPAT_REJECT)
160 * @vol_id: ID of this volume
161 * @lnum: logical eraseblock number
162 * @leb_ver: version of this logical eraseblock (IMPORTANT: obsolete, to be
163 * removed, kept only for not breaking older UBI users)
164 * @data_size: how many bytes of data this logical eraseblock contains
165 * @used_ebs: total number of used logical eraseblocks in this volume
166 * @data_pad: how many bytes at the end of this physical eraseblock are not
167 * used
168 * @data_crc: CRC checksum of the data stored in this logical eraseblock
169 * @padding1: reserved for future, zeroes
170 * @sqnum: sequence number
171 * @padding2: reserved for future, zeroes
172 * @hdr_crc: volume identifier header CRC checksum
173 *
174 * The @sqnum is the value of the global sequence counter at the time when this
175 * VID header was created. The global sequence counter is incremented each time
176 * UBI writes a new VID header to the flash, i.e. when it maps a logical
177 * eraseblock to a new physical eraseblock. The global sequence counter is an
178 * unsigned 64-bit integer and we assume it never overflows. The @sqnum
179 * (sequence number) is used to distinguish between older and newer versions of
180 * logical eraseblocks.
181 *
182 * There are 2 situations when there may be more then one physical eraseblock
183 * corresponding to the same logical eraseblock, i.e., having the same @vol_id
184 * and @lnum values in the volume identifier header. Suppose we have a logical
185 * eraseblock L and it is mapped to the physical eraseblock P.
186 *
187 * 1. Because UBI may erase physical eraseblocks asynchronously, the following
188 * situation is possible: L is asynchronously erased, so P is scheduled for
189 * erasure, then L is written to,i.e. mapped to another physical eraseblock P1,
190 * so P1 is written to, then an unclean reboot happens. Result - there are 2
191 * physical eraseblocks P and P1 corresponding to the same logical eraseblock
192 * L. But P1 has greater sequence number, so UBI picks P1 when it attaches the
193 * flash.
194 *
195 * 2. From time to time UBI moves logical eraseblocks to other physical
196 * eraseblocks for wear-leveling reasons. If, for example, UBI moves L from P
197 * to P1, and an unclean reboot happens before P is physically erased, there
198 * are two physical eraseblocks P and P1 corresponding to L and UBI has to
199 * select one of them when the flash is attached. The @sqnum field says which
200 * PEB is the original (obviously P will have lower @sqnum) and the copy. But
201 * it is not enough to select the physical eraseblock with the higher sequence
202 * number, because the unclean reboot could have happen in the middle of the
203 * copying process, so the data in P is corrupted. It is also not enough to
204 * just select the physical eraseblock with lower sequence number, because the
205 * data there may be old (consider a case if more data was added to P1 after
206 * the copying). Moreover, the unclean reboot may happen when the erasure of P
207 * was just started, so it result in unstable P, which is "mostly" OK, but
208 * still has unstable bits.
209 *
210 * UBI uses the @copy_flag field to indicate that this logical eraseblock is a
211 * copy. UBI also calculates data CRC when the data is moved and stores it at
212 * the @data_crc field of the copy (P1). So when UBI needs to pick one physical
213 * eraseblock of two (P or P1), the @copy_flag of the newer one (P1) is
214 * examined. If it is cleared, the situation* is simple and the newer one is
215 * picked. If it is set, the data CRC of the copy (P1) is examined. If the CRC
216 * checksum is correct, this physical eraseblock is selected (P1). Otherwise
217 * the older one (P) is selected.
218 *
219 * Note, there is an obsolete @leb_ver field which was used instead of @sqnum
220 * in the past. But it is not used anymore and we keep it in order to be able
221 * to deal with old UBI images. It will be removed at some point.
222 *
223 * There are 2 sorts of volumes in UBI: user volumes and internal volumes.
224 * Internal volumes are not seen from outside and are used for various internal
225 * UBI purposes. In this implementation there is only one internal volume - the
226 * layout volume. Internal volumes are the main mechanism of UBI extensions.
227 * For example, in future one may introduce a journal internal volume. Internal
228 * volumes have their own reserved range of IDs.
229 *
230 * The @compat field is only used for internal volumes and contains the "degree
231 * of their compatibility". It is always zero for user volumes. This field
232 * provides a mechanism to introduce UBI extensions and to be still compatible
233 * with older UBI binaries. For example, if someone introduced a journal in
234 * future, he would probably use %UBI_COMPAT_DELETE compatibility for the
235 * journal volume. And in this case, older UBI binaries, which know nothing
236 * about the journal volume, would just delete this volume and work perfectly
237 * fine. This is similar to what Ext2fs does when it is fed by an Ext3fs image
238 * - it just ignores the Ext3fs journal.
239 *
240 * The @data_crc field contains the CRC checksum of the contents of the logical
241 * eraseblock if this is a static volume. In case of dynamic volumes, it does
242 * not contain the CRC checksum as a rule. The only exception is when the
243 * data of the physical eraseblock was moved by the wear-leveling unit, then
244 * the wear-leveling unit calculates the data CRC and stores it in the
245 * @data_crc field. And of course, the @copy_flag is %in this case.
246 *
247 * The @data_size field is used only for static volumes because UBI has to know
248 * how many bytes of data are stored in this eraseblock. For dynamic volumes,
249 * this field usually contains zero. The only exception is when the data of the
250 * physical eraseblock was moved to another physical eraseblock for
251 * wear-leveling reasons. In this case, UBI calculates CRC checksum of the
252 * contents and uses both @data_crc and @data_size fields. In this case, the
253 * @data_size field contains data size.
254 *
255 * The @used_ebs field is used only for static volumes and indicates how many
256 * eraseblocks the data of the volume takes. For dynamic volumes this field is
257 * not used and always contains zero.
258 *
259 * The @data_pad is calculated when volumes are created using the alignment
260 * parameter. So, effectively, the @data_pad field reduces the size of logical
261 * eraseblocks of this volume. This is very handy when one uses block-oriented
262 * software (say, cramfs) on top of the UBI volume.
263 */
264struct ubi_vid_hdr {
265 ubi32_t magic;
266 uint8_t version;
267 uint8_t vol_type;
268 uint8_t copy_flag;
269 uint8_t compat;
270 ubi32_t vol_id;
271 ubi32_t lnum;
272 ubi32_t leb_ver; /* obsolete, to be removed, don't use */
273 ubi32_t data_size;
274 ubi32_t used_ebs;
275 ubi32_t data_pad;
276 ubi32_t data_crc;
277 uint8_t padding1[4];
278 ubi64_t sqnum;
279 uint8_t padding2[12];
280 ubi32_t hdr_crc;
281} __attribute__ ((packed));
282
283/* Internal UBI volumes count */
284#define UBI_INT_VOL_COUNT 1
285
286/*
287 * Starting ID of internal volumes. There is reserved room for 4096 internal
288 * volumes.
289 */
290#define UBI_INTERNAL_VOL_START (0x7FFFFFFF - 4096)
291
292/* The layout volume contains the volume table */
293
294#define UBI_LAYOUT_VOL_ID UBI_INTERNAL_VOL_START
295#define UBI_LAYOUT_VOLUME_EBS 2
296#define UBI_LAYOUT_VOLUME_NAME "layout volume"
297#define UBI_LAYOUT_VOLUME_COMPAT UBI_COMPAT_REJECT
298
299/* The maximum number of volumes per one UBI device */
300#define UBI_MAX_VOLUMES 128
301
302/* The maximum volume name length */
303#define UBI_VOL_NAME_MAX 127
304
305/* Size of the volume table record */
306#define UBI_VTBL_RECORD_SIZE sizeof(struct ubi_vtbl_record)
307
308/* Size of the volume table record without the ending CRC */
309#define UBI_VTBL_RECORD_SIZE_CRC (UBI_VTBL_RECORD_SIZE - sizeof(ubi32_t))
310
311/**
312 * struct ubi_vtbl_record - a record in the volume table.
313 * @reserved_pebs: how many physical eraseblocks are reserved for this volume
314 * @alignment: volume alignment
315 * @data_pad: how many bytes are unused at the end of the each physical
316 * eraseblock to satisfy the requested alignment
317 * @vol_type: volume type (%UBI_DYNAMIC_VOLUME or %UBI_STATIC_VOLUME)
318 * @upd_marker: if volume update was started but not finished
319 * @name_len: volume name length
320 * @name: the volume name
321 * @padding2: reserved, zeroes
322 * @crc: a CRC32 checksum of the record
323 *
324 * The volume table records are stored in the volume table, which is stored in
325 * the layout volume. The layout volume consists of 2 logical eraseblock, each
326 * of which contains a copy of the volume table (i.e., the volume table is
327 * duplicated). The volume table is an array of &struct ubi_vtbl_record
328 * objects indexed by the volume ID.
329 *
330 * If the size of the logical eraseblock is large enough to fit
331 * %UBI_MAX_VOLUMES records, the volume table contains %UBI_MAX_VOLUMES
332 * records. Otherwise, it contains as many records as it can fit (i.e., size of
333 * logical eraseblock divided by sizeof(struct ubi_vtbl_record)).
334 *
335 * The @upd_marker flag is used to implement volume update. It is set to %1
336 * before update and set to %0 after the update. So if the update operation was
337 * interrupted, UBI knows that the volume is corrupted.
338 *
339 * The @alignment field is specified when the volume is created and cannot be
340 * later changed. It may be useful, for example, when a block-oriented file
341 * system works on top of UBI. The @data_pad field is calculated using the
342 * logical eraseblock size and @alignment. The alignment must be multiple to the
343 * minimal flash I/O unit. If @alignment is 1, all the available space of
344 * the physical eraseblocks is used.
345 *
346 * Empty records contain all zeroes and the CRC checksum of those zeroes.
347 */
348struct ubi_vtbl_record {
349 ubi32_t reserved_pebs;
350 ubi32_t alignment;
351 ubi32_t data_pad;
352 uint8_t vol_type;
353 uint8_t upd_marker;
354 ubi16_t name_len;
355 uint8_t name[UBI_VOL_NAME_MAX+1];
356 uint8_t padding2[24];
357 ubi32_t crc;
358} __attribute__ ((packed));
359
360#endif /* !__UBI_HEADER_H__ */
diff --git a/include/mtd/ubi-user.h b/include/mtd/ubi-user.h
new file mode 100644
index 000000000000..fe06ded0e6b8
--- /dev/null
+++ b/include/mtd/ubi-user.h
@@ -0,0 +1,161 @@
1/*
2 * Copyright (c) International Business Machines Corp., 2006
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
12 * the 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 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21#ifndef __UBI_USER_H__
22#define __UBI_USER_H__
23
24/*
25 * UBI volume creation
26 * ~~~~~~~~~~~~~~~~~~~
27 *
28 * UBI volumes are created via the %UBI_IOCMKVOL IOCTL command of UBI character
29 * device. A &struct ubi_mkvol_req object has to be properly filled and a
30 * pointer to it has to be passed to the IOCTL.
31 *
32 * UBI volume deletion
33 * ~~~~~~~~~~~~~~~~~~~
34 *
35 * To delete a volume, the %UBI_IOCRMVOL IOCTL command of the UBI character
36 * device should be used. A pointer to the 32-bit volume ID hast to be passed
37 * to the IOCTL.
38 *
39 * UBI volume re-size
40 * ~~~~~~~~~~~~~~~~~~
41 *
42 * To re-size a volume, the %UBI_IOCRSVOL IOCTL command of the UBI character
43 * device should be used. A &struct ubi_rsvol_req object has to be properly
44 * filled and a pointer to it has to be passed to the IOCTL.
45 *
46 * UBI volume update
47 * ~~~~~~~~~~~~~~~~~
48 *
49 * Volume update should be done via the %UBI_IOCVOLUP IOCTL command of the
50 * corresponding UBI volume character device. A pointer to a 64-bit update
51 * size should be passed to the IOCTL. After then, UBI expects user to write
52 * this number of bytes to the volume character device. The update is finished
53 * when the claimed number of bytes is passed. So, the volume update sequence
54 * is something like:
55 *
56 * fd = open("/dev/my_volume");
57 * ioctl(fd, UBI_IOCVOLUP, &image_size);
58 * write(fd, buf, image_size);
59 * close(fd);
60 */
61
62/*
63 * When a new volume is created, users may either specify the volume number they
64 * want to create or to let UBI automatically assign a volume number using this
65 * constant.
66 */
67#define UBI_VOL_NUM_AUTO (-1)
68
69/* Maximum volume name length */
70#define UBI_MAX_VOLUME_NAME 127
71
72/* IOCTL commands of UBI character devices */
73
74#define UBI_IOC_MAGIC 'o'
75
76/* Create an UBI volume */
77#define UBI_IOCMKVOL _IOW(UBI_IOC_MAGIC, 0, struct ubi_mkvol_req)
78/* Remove an UBI volume */
79#define UBI_IOCRMVOL _IOW(UBI_IOC_MAGIC, 1, int32_t)
80/* Re-size an UBI volume */
81#define UBI_IOCRSVOL _IOW(UBI_IOC_MAGIC, 2, struct ubi_rsvol_req)
82
83/* IOCTL commands of UBI volume character devices */
84
85#define UBI_VOL_IOC_MAGIC 'O'
86
87/* Start UBI volume update */
88#define UBI_IOCVOLUP _IOW(UBI_VOL_IOC_MAGIC, 0, int64_t)
89/* An eraseblock erasure command, used for debugging, disabled by default */
90#define UBI_IOCEBER _IOW(UBI_VOL_IOC_MAGIC, 1, int32_t)
91
92/*
93 * UBI volume type constants.
94 *
95 * @UBI_DYNAMIC_VOLUME: dynamic volume
96 * @UBI_STATIC_VOLUME: static volume
97 */
98enum {
99 UBI_DYNAMIC_VOLUME = 3,
100 UBI_STATIC_VOLUME = 4
101};
102
103/**
104 * struct ubi_mkvol_req - volume description data structure used in
105 * volume creation requests.
106 * @vol_id: volume number
107 * @alignment: volume alignment
108 * @bytes: volume size in bytes
109 * @vol_type: volume type (%UBI_DYNAMIC_VOLUME or %UBI_STATIC_VOLUME)
110 * @padding1: reserved for future, not used
111 * @name_len: volume name length
112 * @padding2: reserved for future, not used
113 * @name: volume name
114 *
115 * This structure is used by userspace programs when creating new volumes. The
116 * @used_bytes field is only necessary when creating static volumes.
117 *
118 * The @alignment field specifies the required alignment of the volume logical
119 * eraseblock. This means, that the size of logical eraseblocks will be aligned
120 * to this number, i.e.,
121 * (UBI device logical eraseblock size) mod (@alignment) = 0.
122 *
123 * To put it differently, the logical eraseblock of this volume may be slightly
124 * shortened in order to make it properly aligned. The alignment has to be
125 * multiple of the flash minimal input/output unit, or %1 to utilize the entire
126 * available space of logical eraseblocks.
127 *
128 * The @alignment field may be useful, for example, when one wants to maintain
129 * a block device on top of an UBI volume. In this case, it is desirable to fit
130 * an integer number of blocks in logical eraseblocks of this UBI volume. With
131 * alignment it is possible to update this volume using plane UBI volume image
132 * BLOBs, without caring about how to properly align them.
133 */
134struct ubi_mkvol_req {
135 int32_t vol_id;
136 int32_t alignment;
137 int64_t bytes;
138 int8_t vol_type;
139 int8_t padding1;
140 int16_t name_len;
141 int8_t padding2[4];
142 char name[UBI_MAX_VOLUME_NAME+1];
143} __attribute__ ((packed));
144
145/**
146 * struct ubi_rsvol_req - a data structure used in volume re-size requests.
147 * @vol_id: ID of the volume to re-size
148 * @bytes: new size of the volume in bytes
149 *
150 * Re-sizing is possible for both dynamic and static volumes. But while dynamic
151 * volumes may be re-sized arbitrarily, static volumes cannot be made to be
152 * smaller then the number of bytes they bear. To arbitrarily shrink a static
153 * volume, it must be wiped out first (by means of volume update operation with
154 * zero number of bytes).
155 */
156struct ubi_rsvol_req {
157 int64_t bytes;
158 int32_t vol_id;
159} __attribute__ ((packed));
160
161#endif /* __UBI_USER_H__ */