码迷,mamicode.com
首页 > 其他好文 > 详细

ubi使用(转)

时间:2016-07-15 19:38:40      阅读:361      评论:0      收藏:0      [点我收藏+]

标签:

1. ubifs号称性能比yaffs2 好,同时压缩可读写,文件系统image体较小同时可写,相当不错

2. ubifs制作

(1) mkfs.ubifs

mkfs.ubifs -r system -m 2048 -e 126976 -c 1057 -x zlib -o system.ubifs

-r说的目录 

-m说的是nand flash的页大小,一般都是2048或者4096,当然也有些是512,硬件特性决定好了,或者说驱动规定好了,有些平台4096页当做2048用也是用的

-e说的是逻辑擦除块大小,大家知道nand flash页读页写块擦,一个设备多个块,一个块多个页,一般也都是一个块是64个页,这样算一下无论擦除块大小就是2048*64=131072,-e的算法是物理擦除块大小-2*页大小,这里就是131072-2*2048=126976

-c说的是最大逻辑块编号,这个很重要,不能大也不能小,太小也要大于image大小,太大mount有问题,计算起点是分区的物理块数量,比如128MiB的mtd分区,物理块数量是128MiB/2048/64 = 1024个,需要减去2个坏块保留块,减去1个wear-leveling块,还要减去1个eba的块,等等,比如最终的值是1022,注意,如果物理上这个分区有坏块的话,kernel会扫描到的,这时候,我们计算的这个值就要减去坏块数了,否则会有逻辑块大于物理块数的内核问题mount失败,确切知道坏块数是比较困难的,一般做法是做一个坏块容忍数,比如20个,这样我们再减去20个坏块,不要担心这个会浪费空间,ubinize的autoresize选项就是解决这个问题的。具体的这个值需要计算。!!!!

-o说的image名字 

-x说的是压缩方法,默认是lzo,还支持zlib,zlib压缩率高些,但是lzo压缩解压速度快 

(2)  ubinize

 

ubinize -o system.ubi -m 2048 -p 131072 ubinize.cfg

-o说的是输出image

-m还是页大小

-p是物理擦除块大小

ubinize.cfg是volume配置文件,例子如下:

 

  1. [ubifs]
  2. mode=ubi
  3. image=system.ubifs #说的是mkfs.ubifs的结果
  4. vol_id=0
  5. vol_size=100MiB #说的是volume大小,用-e和-c的值做乘法计算,一般不用写,autoresize会自动根据mtd分区大小适应,默认值是image大小,写了这个作用是帮助检查image是否超过了分区限制,制作时候就提示,否则mount会出错。-c的值是经过计算的最大值了,不过autoresize参数会自适应大小,不会浪费空间的。
  6. vol_type=dynamic
  7. vol_alignment=1
  8. vol_name=system #说的是分区名字
  9. vol_flags=autoresize

 

 

(4) uboot支持

-0-  需要打开的配置 - 需要烧写mkfs.ubifs结果时候需要打开,ubinize处理过的不需要

CONFIG_CMD_UBI

CONFIG_CMD_UBIFS

CONFIG_LZO

CONFIG_RBTREE

CONFIG_ZLIB

CONFIG_GZIP 

-1- ubifs烧写

mkfs.ubifs工具制作的结果,就是ubifs镜像,不包含volume信息,需要用

nand erase.part system

ubi part system - 激活分区

ubi create system - 创建分区 

ubi write 84000000 system xxxxx - xxxxx表示镜像实际大小

-2- ubi volume bin烧写

 mkfs.ubifs后,使用ubinize处理了ubifs的镜像后,镜像含有了volume信息,直接

nand write 84000000 system xxxxx - xxxxx表示镜像实际大小

(5) kernel支持

 

  • Device Drivers --->
  •     Memory Technology Device(MTD) support --->
  •       UBI - Unsorted block images --->
  •         <*> Enable UBI - Unsorted block images
  • File systems --->
  •     Miscellaneous filesystems --->
  •         <*> UBIFS file system support
  •      [*] Advanced compression options
  •         [*] LZO compression support
  •         [*] ZLIB compression support

 

(6) android支持 

-1- andriod/system/core/rootdir/init.rc

mount yaffs2 mtd@userdata /data nosuid nodev

改为:

mount ubifs ubi@userdata /data nosuid nodev

-2- 增加对ubifs的mount支持

 

  • andriod/system/core/init/builtins.c 
  • 在 } else if (!strncmp(source, "loop@", 5)) {
  • 之前加上
  • + }else if (!strncmp(source, "ubi@", 4)) {
  • + n = ubi_attach_mtd(source + 4);
  • + if (n < 0) {
  • + return -1;
  • + }
  • + sprintf(tmp, "/dev/ubi%d_0", n);
  • + if (wait)
  • + wait_for_file(tmp, COMMAND_RETRY_TIMEOUT);
  • + if (mount(tmp, target, system, flags, options) < 0) {
  • + ubi_detach_dev(n);
  • + return -1;
  • + }
  • + return 0;
  • + }else if (!strncmp(source, "loop@", 5)) {
  • 3. andriod/system/core/init/init.c
  • 在static int property_triggers_enabled = 0;之后加上
  • +static unsigned ubifs_enabled = 1;
  • static int set_init_properties_action(int nargs, char **args)
  • {
  •     property_set("ro.revision", tmp);
  • + property_set("ro.ubifs",ubifs_enabled ? "1" : "0");
  •     return 0;
  • }
  • int main(int argc, char **argv)
  • {
  •    action_for_each_trigger("post-fs", action_add_queue_tail);
  • + action_for_each_trigger("ubi-fs", action_add_queue_tail);
  • }
  • 4. andriod/system/core/init/util.h
  • void get_hardware_name(char *hardware, unsigned int *revision);
  • +int ubi_attach_mtd(const char *name);
  • +int ubi_detach_dev(int dev);
  • 5. andriod/system/core/init/util.c
  • #include <sys/un.h>
  • +#include <sys/ioctl.h>
  • #include "util.h"
  • +#include "ubi-user.h"
  • +#define UBI_CTRL_DEV "/dev/ubi_ctrl"
  • +#define UBI_SYS_PATH "/sys/class/ubi"
  • 在最后添加下面三个函数
  • static int ubi_dev_read_int(int dev, const char *file, int def)
  • {
  •     int fd, val = def;
  •     char path[128], buf[64];
  •     sprintf(path, UBI_SYS_PATH "/ubi%d/%s", dev, file);
  •     wait_for_file(path, 5);
  •     fd = open(path, O_RDONLY);
  •     if (fd == -1) {
  •         return val;
  •     }
  •     if (read(fd, buf, 64) > 0) {
  •         val = atoi(buf);
  •     }
  •     close(fd);
  •     return val;
  • }
  • int ubi_attach_mtd(const char *name)
  • {
  •     int ret;
  •     int mtd_num, ubi_num;
  •     int ubi_ctrl, ubi_dev;
  •     int vols, avail_lebs, leb_size;
  •     char path[128];
  •     struct ubi_attach_req attach_req;
  •     struct ubi_mkvol_req mkvol_req;
  •     mtd_num = mtd_name_to_number(name);
  •     if (mtd_num == -1) {
  •         return -1;
  •     }
  •     ubi_ctrl = open(UBI_CTRL_DEV, O_RDONLY);
  •     if (ubi_ctrl == -1) {
  •         return -1;
  •     }
  •     memset(&attach_req, 0, sizeof(struct ubi_attach_req));
  •     attach_req.ubi_num = UBI_DEV_NUM_AUTO;
  •     attach_req.mtd_num = mtd_num;
  •     ret = ioctl(ubi_ctrl, UBI_IOCATT, &attach_req);
  •     if (ret == -1) {
  •         close(ubi_ctrl);
  •         return -1;
  •     }
  •     ubi_num = attach_req.ubi_num;
  •     vols = ubi_dev_read_int(ubi_num, "volumes_count", -1);
  •     if (vols == 0) {
  •         sprintf(path, "/dev/ubi%d", ubi_num);
  •         ubi_dev = open(path, O_RDONLY);
  •         if (ubi_dev == -1) {
  •             close(ubi_ctrl);
  •             return ubi_num;
  •         }
  •         avail_lebs = ubi_dev_read_int(ubi_num, "avail_eraseblocks", 0);
  •         leb_size = ubi_dev_read_int(ubi_num, "eraseblock_size", 0);
  •         memset(&mkvol_req, 0, sizeof(struct ubi_mkvol_req));
  •         mkvol_req.vol_id = UBI_VOL_NUM_AUTO;
  •         mkvol_req.alignment = 1;
  •         mkvol_req.bytes = (long long)avail_lebs * leb_size;
  •         mkvol_req.vol_type = UBI_DYNAMIC_VOLUME;
  •         ret = snprintf(mkvol_req.name, UBI_MAX_VOLUME_NAME + 1, "%s", name);
  •         mkvol_req.name_len = ret;
  •         ioctl(ubi_dev, UBI_IOCMKVOL, &mkvol_req);
  •         close(ubi_dev);
  •     }
  •     close(ubi_ctrl);
  •     return ubi_num;
  • }
  • int ubi_detach_dev(int dev)
  • {
  •     int ret, ubi_ctrl;
  •     ubi_ctrl = open(UBI_CTRL_DEV, O_RDONLY);
  •     if (ubi_ctrl == -1) {
  •         return -1;
  •     }
  •     ret = ioctl(ubi_ctrl, UBI_IOCDET, &dev);
  •     close(ubi_ctrl);
  •     return ret;
  • }
  • 6.增加文件
  • andriod/system/core/init/ubi-user.h
  • /*
  •  * Copyright (c) International Business Machines Corp., 2006
  •  *
  •  * This program is free software; you can redistribute it and/or modify
  •  * it under the terms of the GNU General Public License as published by
  •  * the Free Software Foundation; either version 2 of the License, or
  •  * (at your option) any later version.
  •  *
  •  * This program is distributed in the hope that it will be useful,
  •  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  •  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
  •  * the GNU General Public License for more details.
  •  *
  •  * You should have received a copy of the GNU General Public License
  •  * along with this program; if not, write to the Free Software
  •  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  •  *
  •  * Author: Artem Bityutskiy (Битюцкий Артём)
  •  */
  • #ifndef __UBI_USER_H__
  • #define __UBI_USER_H__
  • /*
  •  * UBI device creation (the same as MTD device attachment)
  •  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  •  *
  •  * MTD devices may be attached using %UBI_IOCATT ioctl command of the UBI
  •  * control device. The caller has to properly fill and pass
  •  * &struct ubi_attach_req object - UBI will attach the MTD device specified in
  •  * the request and return the newly created UBI device number as the ioctl
  •  * return value.
  •  *
  •  * UBI device deletion (the same as MTD device detachment)
  •  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  •  *
  •  * An UBI device maybe deleted with %UBI_IOCDET ioctl command of the UBI
  •  * control device.
  •  *
  •  * UBI volume creation
  •  * ~~~~~~~~~~~~~~~~~~~
  •  *
  •  * UBI volumes are created via the %UBI_IOCMKVOL IOCTL command of UBI character
  •  * device. A &struct ubi_mkvol_req object has to be properly filled and a
  •  * pointer to it has to be passed to the IOCTL.
  •  *
  •  * UBI volume deletion
  •  * ~~~~~~~~~~~~~~~~~~~
  •  *
  •  * To delete a volume, the %UBI_IOCRMVOL IOCTL command of the UBI character
  •  * device should be used. A pointer to the 32-bit volume ID hast to be passed
  •  * to the IOCTL.
  •  *
  •  * UBI volume re-size
  •  * ~~~~~~~~~~~~~~~~~~
  •  *
  •  * To re-size a volume, the %UBI_IOCRSVOL IOCTL command of the UBI character
  •  * device should be used. A &struct ubi_rsvol_req object has to be properly
  •  * filled and a pointer to it has to be passed to the IOCTL.
  •  *
  •  * UBI volume update
  •  * ~~~~~~~~~~~~~~~~~
  •  *
  •  * Volume update should be done via the %UBI_IOCVOLUP IOCTL command of the
  •  * corresponding UBI volume character device. A pointer to a 64-bit update
  •  * size should be passed to the IOCTL. After this, UBI expects user to write
  •  * this number of bytes to the volume character device. The update is finished
  •  * when the claimed number of bytes is passed. So, the volume update sequence
  •  * is something like:
  •  *
  •  * fd = open("/dev/my_volume");
  •  * ioctl(fd, UBI_IOCVOLUP, &image_size);
  •  * write(fd, buf, image_size);
  •  * close(fd);
  •  *
  •  * Atomic eraseblock change
  •  * ~~~~~~~~~~~~~~~~~~~~~~~~
  •  *
  •  * Atomic eraseblock change operation is done via the %UBI_IOCEBCH IOCTL
  •  * command of the corresponding UBI volume character device. A pointer to
  •  * &struct ubi_leb_change_req has to be passed to the IOCTL. Then the user is
  •  * expected to write the requested amount of bytes. This is similar to the
  •  * "volume update" IOCTL.
  •  */
  • /*
  •  * When a new UBI volume or UBI device is created, users may either specify the
  •  * volume/device number they want to create or to let UBI automatically assign
  •  * the number using these constants.
  •  */
  • #define UBI_VOL_NUM_AUTO (-1)
  • #define UBI_DEV_NUM_AUTO (-1)
  • /* Maximum volume name length */
  • #define UBI_MAX_VOLUME_NAME 127
  • /* IOCTL commands of UBI character devices */
  • #define UBI_IOC_MAGIC ‘o‘
  • /* Create an UBI volume */
  • #define UBI_IOCMKVOL _IOW(UBI_IOC_MAGIC, 0, struct ubi_mkvol_req)
  • /* Remove an UBI volume */
  • #define UBI_IOCRMVOL _IOW(UBI_IOC_MAGIC, 1, int32_t)
  • /* Re-size an UBI volume */
  • #define UBI_IOCRSVOL _IOW(UBI_IOC_MAGIC, 2, struct ubi_rsvol_req)
  • /* IOCTL commands of the UBI control character device */
  • #define UBI_CTRL_IOC_MAGIC ‘o‘
  • /* Attach an MTD device */
  • #define UBI_IOCATT _IOW(UBI_CTRL_IOC_MAGIC, 64, struct ubi_attach_req)
  • /* Detach an MTD device */
  • #define UBI_IOCDET _IOW(UBI_CTRL_IOC_MAGIC, 65, int32_t)
  • /* IOCTL commands of UBI volume character devices */
  • #define UBI_VOL_IOC_MAGIC ‘O‘
  • /* Start UBI volume update */
  • #define UBI_IOCVOLUP _IOW(UBI_VOL_IOC_MAGIC, 0, int64_t)
  • /* An eraseblock erasure command, used for debugging, disabled by default */
  • #define UBI_IOCEBER _IOW(UBI_VOL_IOC_MAGIC, 1, int32_t)
  • /* An atomic eraseblock change command */
  • #define UBI_IOCEBCH _IOW(UBI_VOL_IOC_MAGIC, 2, int32_t)
  • /* Maximum MTD device name length supported by UBI */
  • #define MAX_UBI_MTD_NAME_LEN 127
  • /*
  •  * UBI data type hint constants.
  •  *
  •  * UBI_LONGTERM: long-term data
  •  * UBI_SHORTTERM: short-term data
  •  * UBI_UNKNOWN: data persistence is unknown
  •  *
  •  * These constants are used when data is written to UBI volumes in order to
  •  * help the UBI wear-leveling unit to find more appropriate physical
  •  * eraseblocks.
  •  */
  • enum {
  •     UBI_LONGTERM = 1,
  •     UBI_SHORTTERM = 2,
  •     UBI_UNKNOWN = 3,
  • };
  • /*
  •  * UBI volume type constants.
  •  *
  •  * @UBI_DYNAMIC_VOLUME: dynamic volume
  •  * @UBI_STATIC_VOLUME: static volume
  •  */
  • enum {
  •     UBI_DYNAMIC_VOLUME = 3,
  •     UBI_STATIC_VOLUME = 4,
  • };
  • /**
  •  * struct ubi_attach_req - attach MTD device request.
  •  * @ubi_num: UBI device number to create
  •  * @mtd_num: MTD device number to attach
  •  * @vid_hdr_offset: VID header offset (use defaults if %0)
  •  * @padding: reserved for future, not used, has to be zeroed
  •  *
  •  * This data structure is used to specify MTD device UBI has to attach and the
  •  * parameters it has to use. The number which should be assigned to the new UBI
  •  * device is passed in @ubi_num. UBI may automatically assign the number if
  •  * @UBI_DEV_NUM_AUTO is passed. In this case, the device number is returned in
  •  * @ubi_num.
  •  *
  •  * Most applications should pass %0 in @vid_hdr_offset to make UBI use default
  •  * offset of the VID header within physical eraseblocks. The default offset is
  •  * the next min. I/O unit after the EC header. For example, it will be offset
  •  * 512 in case of a 512 bytes page NAND flash with no sub-page support. Or
  •  * it will be 512 in case of a 2KiB page NAND flash with 4 512-byte sub-pages.
  •  *
  •  * But in rare cases, if this optimizes things, the VID header may be placed to
  •  * a different offset. For example, the boot-loader might do things faster if the
  •  * VID header sits at the end of the first 2KiB NAND page with 4 sub-pages. As
  •  * the boot-loader would not normally need to read EC headers (unless it needs
  •  * UBI in RW mode), it might be faster to calculate ECC. This is weird example,
  •  * but it real-life example. So, in this example, @vid_hdr_offer would be
  •  * 2KiB-64 bytes = 1984. Note, that this position is not even 512-bytes
  •  * aligned, which is OK, as UBI is clever enough to realize this is 4th sub-page
  •  * of the first page and add needed padding.
  •  */
  • struct ubi_attach_req {
  •     int32_t ubi_num;
  •     int32_t mtd_num;
  •     int32_t vid_hdr_offset;
  •     uint8_t padding[12];
  • };
  • /**
  •  * struct ubi_mkvol_req - volume description data structure used in
  •  * volume creation requests.
  •  * @vol_id: volume number
  •  * @alignment: volume alignment
  •  * @bytes: volume size in bytes
  •  * @vol_type: volume type (%UBI_DYNAMIC_VOLUME or %UBI_STATIC_VOLUME)
  •  * @padding1: reserved for future, not used, has to be zeroed
  •  * @name_len: volume name length
  •  * @padding2: reserved for future, not used, has to be zeroed
  •  * @name: volume name
  •  *
  •  * This structure is used by user-space programs when creating new volumes. The
  •  * @used_bytes field is only necessary when creating static volumes.
  •  *
  •  * The @alignment field specifies the required alignment of the volume logical
  •  * eraseblock. This means, that the size of logical eraseblocks will be aligned
  •  * to this number, i.e.,
  •  *    (UBI device logical eraseblock size) mod (@alignment) = 0.
  •  *
  •  * To put it differently, the logical eraseblock of this volume may be slightly
  •  * shortened in order to make it properly aligned. The alignment has to be
  •  * multiple of the flash minimal input/output unit, or %1 to utilize the entire
  •  * available space of logical eraseblocks.
  •  *
  •  * The @alignment field may be useful, for example, when one wants to maintain
  •  * a block device on top of an UBI volume. In this case, it is desirable to fit
  •  * an integer number of blocks in logical eraseblocks of this UBI volume. With
  •  * alignment it is possible to update this volume using plane UBI volume image
  •  * BLOBs, without caring about how to properly align them.
  •  */
  • struct ubi_mkvol_req {
  •     int32_t vol_id;
  •     int32_t alignment;
  •     int64_t bytes;
  •     int8_t vol_type;
  •     int8_t padding1;
  •     int16_t name_len;
  •     int8_t padding2[4];
  •     char name[UBI_MAX_VOLUME_NAME + 1];
  • } __attribute__ ((packed));
  • /**
  •  * struct ubi_rsvol_req - a data structure used in volume re-size requests.
  •  * @vol_id: ID of the volume to re-size
  •  * @bytes: new size of the volume in bytes
  •  *
  •  * Re-sizing is possible for both dynamic and static volumes. But while dynamic
  •  * volumes may be re-sized arbitrarily, static volumes cannot be made to be
  •  * smaller then the number of bytes they bear. To arbitrarily shrink a static
  •  * volume, it must be wiped out first (by means of volume update operation with
  •  * zero number of bytes).
  •  */
  • struct ubi_rsvol_req {
  •     int64_t bytes;
  •     int32_t vol_id;
  • } __attribute__ ((packed));
  • /**
  •  * struct ubi_leb_change_req - a data structure used in atomic logical
  •  * eraseblock change requests.
  •  * @lnum: logical eraseblock number to change
  •  * @bytes: how many bytes will be written to the logical eraseblock
  •  * @dtype: data type (%UBI_LONGTERM, %UBI_SHORTTERM, %UBI_UNKNOWN)
  •  * @padding: reserved for future, not used, has to be zeroed
  •  */
  • struct ubi_leb_change_req {
  •     int32_t lnum;
  •     int32_t bytes;
  •     uint8_t dtype;
  •     uint8_t padding[7];
  • } __attribute__ ((packed));
  • #endif /* __UBI_USER_H__ */

 

 

 

3. 常见问题

(1) ubifs_check_node: bad CRC: calculated 0xca82b3d7, read 0x9be0e26

 

ubifs_check_node: bad node at LEB 51:45312

 

注意这个crc错误,说的错误并不是LED 0的问题,也就是说第一个逻辑块没问题,CRC是正确的,这个问题的一个解决办法是要精确mkfs.ubifs时候的-c参数的值,不能太大了 

(2)

[    5.433349] UBIFS error (pid 71): ubifs_read_node: bad node type (150 but expected 1)
[    5.434204] UBIFS error (pid 71): ubifs_read_node: bad node at LEB 524:4072, LEB mapping status 1
[    5.435241] Not a node, first 24 bytes:
[    5.435729] 00000000: 34 fb 21 ee 84 18 69 2d 60 b0 33 e6 74 f8 1c 15 da ca a1 c9 96 e3 ac 51                          4.!...i-`.3.t..........Q

 

 

这个问题是,ubifs给nand驱动的buffer不一定是按照硬件对齐要求的,所以要驱动来判断,dma等地址需要页对齐等特性 

(3) crc错误,但是LED 0:0

可能镜像没烧对,或者mtd-utils的版本和内核版本相差较远

(4) 可以擦掉一个分区,并不需要非要烧写image到那个分区,也可以直接mount的,这样可以做实验验证ubi的性能,前提是kernel配置好了,mount正确了

(5) 如果是跟文件系统或者要手动mount,以下步骤供参考:

uboot里, mtd命令后,看到需要用ubifs的mtd分区的编号,比如

 

device nand0 <rda_nand>, # parts = 10
 #: name                size            offset          mask_flags
 0: bootloader          0x00200000      0x00000000      0
 1: boot                0x00800000      0x00800000      0
 2: system              0x08000000      0x02000000      0

 system分区mtd编号是2,命令如下:

 

 nand erase.part system

命令行参数增加

ubi.mtd=2,如果有多个,可以增加,如ubi.mtd=2,ubi.mtd=1,这样传递后,内核启动后会做attach的操作,类似android的init中的attach那样,attach之后,/dev/下就建立好设备了,进入系统后用:

内核控制台里,mount -t ubifs /dev/ubi0_0 /mnt 即可mount 

 

转自:http://www.cnblogs.com/linucos/p/3279381.html

ubi使用(转)

标签:

原文地址:http://www.cnblogs.com/embedded-linux/p/5674218.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!