移植ubifs到dm368成功

我在用TI的dm368開發板，kernel是2.6.32.17，預設的flash檔案系統是jffs2，但是jffs2在大分區下，mount速度很慢，而且佔用ram較多，因此，我想使用ubifs看看效能是否會更好些。

ubifs的原理和配置過程，很多網頁都有介紹的，我給一個連結，大家可以看看，我就不轉載了，我重點說我移植過程中遇到並解決的問題。

http://bbs.chinaunix.net/thread-1954881-1-1.html

 

kerne的配置很簡單，2.6.32裡面都有，選上並重新編譯就好了。

ubiattach、ubimkvol等工具，TI的dvsdk裡面內建了，能用，我就偷了個懶，沒有重新編譯。

 

很輕鬆的照著網頁說明操作了下，mount 分區也成功了，複製檔案也成功了，很高興：）

ubiattach /dev/ubi_ctrl -m 3

ubimkvol  /dev/ubi0 -N rootfs -s 480MiB

mount -t ubifs  ubi0_0 /mnt/nand/ -o sync

但很快就遇到麻煩了，開發板關機重新啟動，我再mout ubi檔案系統就出錯了，提示了一堆錯誤，而且分區是空的，之前複製的檔案不見了。

問題如下：

ubiattach /dev/ubi_ctrl -m 3

UBI error: ubi_io_read: error -74while reading 64 bytes from PEB 1662:0, read 64 bytes

UBI error: ubi_io_read: error -74while reading 64 bytes from PEB 1663:0, read 64 bytes

 

為了分析問題，我把mtd和ubifs的debug log都開啟了，看到了一些資訊；

#define   EBADMSG           74   /* Not a data message */

              /*Nand returns -EBADMSG on ecc errors, but it returns

               * the data. For our userspace tools it isimportant

               * to dump areas with ecc errors !

               * For kernel internal usage it also mightreturn -EUCLEAN

               * to signal the caller that a bitflip hasoccured and has

               * been corrected by the ECC algorithm.

               * Userspace software which accesses NAND thisway

               * must be aware of the fact that it deals withNAND

               */

nand_do_read_ops

       stats= mtd->ecc_stats;

       if(mtd->ecc_stats.failed - stats.failed)

              return-EBADMSG;

  

nand_read_page_hwecc

              stat= chip->ecc.correct(mtd, p, &ecc_code[i], &ecc_calc[i]);

              if(stat < 0)

                     mtd->ecc_stats.failed++;

              else

                     mtd->ecc_stats.corrected+= stat;

看樣子，應該是ECC出錯了，可是我用的是硬體ECC校正，怎麼可能幾乎全部的page都出現ecc校正錯誤了呢？

 

root@dm368-evm:~#flash_eraseall  /dev/mtd3

root@dm368-evm:~# ubiattach/dev/ubi_ctrl -m 3

UBI: attaching mtd3 to ubi0

UBI: physical eraseblock size:   131072 bytes (128 KiB)

UBI: logical eraseblock size:    129024 bytes

UBI: smallest flash I/O unit:    2048

UBI: sub-page size:              512

UBI: VID header offset:          512 (aligned 512)

UBI: data offset:                2048

 

分析下UBI寫資料的過程，

UBI DBG (pid 1484): ubi_io_write:write 512 bytes to PEB 57:0

int ubi_io_write(struct ubi_device*ubi, const void *buf, int pnum, int offset,

               int len)

       addr= (loff_t)pnum * ubi->peb_size + offset;

       err= ubi->mtd->write(ubi->mtd, addr, len, &written, buf);

在io_init函數中可以看到幾個變數的賦值；

       ubi->peb_size   = ubi->mtd->erasesize;

       ubi->peb_count  = mtd_div_by_eb(ubi->mtd->size,ubi->mtd);

       ubi->flash_size= ubi->mtd->size;

從debug log來看，這裡一次寫了512位元組，從某個block的起始page開始，offset是0；

那麼到了nand mtd底層驅動，行為就是；

nand_write->nand_do_write_ops:

       subpage= column || (writelen & (mtd->writesize - 1));

這裡，subpage就是writelen，512；

也可以得知一點，ubifs沒有使用oob，這跟jffs2和yaffs2是不同的；

              /*Partial page write ? */

              if(unlikely(column || writelen < (mtd->writesize - 1))) {

                     cached= 0;

                     bytes= min_t(int, bytes - column, (int) writelen);

                     chip->pagebuf= -1;

                     memset(chip->buffers->databuf,0xff, mtd->writesize);

                     memcpy(&chip->buffers->databuf[column],buf, bytes);

                     wbuf= chip->buffers->databuf;

              }

              ret= chip->write_page(mtd, chip, wbuf, page, cached,

                                   (ops->mode == MTD_OOB_RAW));

下面是write_page函數的代碼；

static int nand_write_page(structmtd_info *mtd, struct nand_chip *chip,

                        const uint8_t *buf, int page, int cached,int raw)

{

       intstatus;

 

       chip->cmdfunc(mtd,NAND_CMD_SEQIN, 0x00, page);

 

       if(unlikely(raw))

              chip->ecc.write_page_raw(mtd,chip, buf);

       else

              chip->ecc.write_page(mtd,chip, buf);

 

       /*

        * Cached progamming disabled for now, Not sureif its worth the

        * trouble. The speed gain is not veryimpressive. (2.3->2.6Mib/s)

        */

       cached= 0;

這裡需要注意的就是raw，如果是MTD_OOB_RAW，那麼不會做ECC校正，也不會把ECC碼寫入OOB；

如果是這樣，在read的時候也必須指定是MTD_OOB_RAW，不需要ECC校正；否則，就會出現我們最開始看到的錯誤；

       if(mtd->ecc_stats.failed - stats.failed)

              return-EBADMSG;

 

那麼，從這裡的情況來看，我們可能已經找到出錯的原因了；ubi使用了subpage write，而底層nand flash驅動實際上是不支援subpage write的，儘管ubi一次唯寫了512位元組，但這個page的其他部分已經不能再次寫入新的資料了。

 

從Nand_base.c(drivers\mtd\nand)來看，large page的nand flash，對subpage write的支援是不完善的，限制條件比較多，比如，不能是MLC的nand flash，不能用硬體ECC；

更嚴重的問題是代碼存在缺陷，在寫入部分data的時候，將其他部分的資料填充為0xff了，然後write整個page，並寫入全部ecc碼到oob，也許這就是前面ecc校正出錯的原因吧。

nand_do_write_ops()

              /*Partial page write ? */

memset(chip->buffers->databuf, 0xff, mtd->writesize);

                     memcpy(&chip->buffers->databuf[column],buf, bytes);

因此，我想到的解決辦法就是在nand驅動中禁止subpage write。

第一步，在chip options中增加NAND_NO_SUBPAGE_WRITE；

static struct davinci_nand_pdatadavinci_nand_data = {

       .options         = NAND_USE_FLASH_BBT|NAND_NO_SUBPAGE_WRITE,

然後重新編譯下載kernel，但問題依舊；

root@dm368-evm:~# ubiattach/dev/ubi_ctrl -m 3

UBI: attaching mtd3 to ubi0

UBI: physical eraseblock size:   131072 bytes (128 KiB)

UBI: logical eraseblock size:    129024 bytes

UBI: smallest flash I/O unit:    2048

UBI: sub-page size:              512

UBI: VID header offset:          512 (aligned 512)

心想很奇怪，為什麼 sub-page size還是512？

回頭查看代碼，想看看sub page size是怎樣計算出來的，

 if (!(chip->options & NAND_NO_SUBPAGE_WRITE) &&
     !(chip->cellinfo & NAND_CI_CELLTYPE_MSK)) {
  switch(chip->ecc.steps) {
  case 2:
   mtd->subpage_sft = 1;
   break;
  case 4:
  case 8:
  case 16:
   mtd->subpage_sft = 2;
   break;
  }
 }
 chip->subpagesize = mtd->writesize >> mtd->subpage_sft;

可是我已經給options增加了NAND_NO_SUBPAGE_WRITE啊？有些懷疑，我就在此處加了列印log，果然是這裡出了問題，

chip->options = 0x10101.

mtd->subpage_sft = 0.

chip->subpagesize = 512.

 

#define NAND_NO_SUBPAGE_WRITE    0x00000200

可是，這裡的options 明明是不對的！那我設定的NAND_NO_SUBPAGE_WRITE在哪裡丟掉了？

 

在下面的函數中有對chip->options賦值改變；

nand_get_flash_type()

              printk(KERN_INFO"nand_get_flash_type 1, chip->options = 0x%x.\n",chip->options);

 

       /*Get chip options, preserve non chip based options */

       chip->options &= ~NAND_CHIPOPTIONS_MSK;

       printk(KERN_INFO"nand_get_flash_type 2, chip->options = 0x%x.\n",chip->options);   

       chip->options|= type->options & NAND_CHIPOPTIONS_MSK;

       printk(KERN_INFO"nand_get_flash_type 3, chip->options = 0x%x.\n",chip->options);

 

       /*

        * Set chip as a default. Board drivers canoverride it, if necessary

        */

       chip->options|= NAND_NO_AUTOINCR;

 

       /*Check if chip is a not a samsung device. Do not clear the

        * options for chips which are not having anextended id.

        */

       if(*maf_id != NAND_MFR_SAMSUNG && !type->pagesize)

              chip->options&= ~NAND_SAMSUNG_LP_OPTIONS;

       printk(KERN_INFO"nand_get_flash_type 4, chip->options = 0x%x.\n",chip->options);

 

nand_get_flash_type 1,chip->options = 0x10200.

nand_get_flash_type 2,chip->options = 0x10000.

nand_get_flash_type 3,chip->options = 0x1011c.

nand_get_flash_type 4,chip->options = 0x10101.

 

問題出在這裡，紅色的代碼！

/* Mask to zero out thechip options, which come from the id table */

#defineNAND_CHIPOPTIONS_MSK (0x0000ffff &~NAND_NO_AUTOINCR)

/* Chip can not autoincrement pages */

#define NAND_NO_AUTOINCR  0x00000001

 

找到問題了，解決辦法就有了，注釋掉紅色的那行代碼，就是他把NAND_NO_SUBPAGE_WRITE給丟掉了。

//chip->options &=~NAND_CHIPOPTIONS_MSK;

再重新編譯下載kernel，問題搞定了！

再次啟動開發板，載入ubi，一切正常了。

root@dm368-evm:/# ubiattach/dev/ubi_ctrl -m 3

UBI: attaching mtd3 to ubi0

UBI: physical eraseblock size:   131072 bytes (128 KiB)

UBI: logical eraseblock size:    126976 bytes

UBI: smallest flash I/O unit:    2048

UBI: VID header offset:          2048 (aligned 2048)

UBI: data offset:                4096

UBI: attached mtd3 to ubi0

UBI: MTD device name:            "filesystem1"

UBI: MTD device size:            512 MiB

UBI: number of good PEBs:        4096

UBI: number of bad PEBs:         0

UBI: max. allowed volumes:       128

UBI: wear-leveling threshold:    4096

UBI: number of internal volumes: 1

UBI: number of user volumes:     1

UBI: available PEBs:             3639

UBI: total number of reserved PEBs:457

UBI: number of PEBs reserved for badPEB handling: 40

UBI: max/mean erase counter: 2/1

UBI: image sequence number: 0

UBI: background thread"ubi_bgt0d" started, PID 1483

UBI device number 0, total 4096 LEBs(520093696 bytes, 496.0 MiB), available 3639 LEBs (462065664 bytes, 440.7 MiB),LEB size 126976 bytes (124.0 KiB)

 

root@dm368-evm:/# mount -tubifs  ubi0_0 /mnt/nand/ -o sync

UBIFS: mounted UBI device 0, volume0, name "rootfs"

UBIFS: file system size:   51171328 bytes (49972 KiB, 48 MiB, 403 LEBs)

UBIFS: journal size:       2539520 bytes (2480 KiB, 2 MiB, 20 LEBs)

UBIFS: media format:       w4/r0 (latest is w4/r0)

UBIFS: default compressor: lzo

UBIFS: reserved for root:  2416947 bytes (2360 KiB)

root@dm368-evm:/# df

Filesystem           1K-blocks      Used Available Use% Mounted on

/dev/root             39544232  16239820 21295632  43% /

none                      1024        24     1000   2% /dev

tmpfs                    16384        20    16364   0% /var/volatile

tmpfs                    21760         0    21760   0% /dev/shm

tmpfs                    16384         0    16384   0% /media/ram

ubi0_0                   45528      1528    41640   4% /mnt/nand

 

最後，總結下，就是要禁止nand flash驅動裡面的subpage write，讓 chip->subpagesize == mtd->writesize 就好了。

這是我的解決辦法。如果有人要使用subpage write，那麼一定要保證你的讀寫邏輯是正確的，要麼不用ecc，要麼各個subpage的ecc都應是正確的。