ubifs烧写遇到《UBI error: ubi_io_read: error -1 while reading 64 bytes from PEB》

终于移植好了2.6.39的内核，烧写ubi文件系统后，第一次打印到/init初始化了，虽然oops了，但还是感觉离最后的成功越来越近了。

可是，偶然断电重启发现，debug口打印却和第一次不一样了，在attach mtd的时候，内核已经开始打印了一些错误，如下

3 cmdlinepart partitions found on MTD device atmel_nand
Creating 3 MTD partitions on "atmel_nand":
0x000000000000-0x000000500000 : "bstrap/uboot/kernel"
0x000000500000-0x000002500000 : "rootfs"
0x000002500000-0x000008000000 : "usrdata"
UBI: attaching mtd1 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
atmel_nand atmel_nand: Too many error.
atmel_nand atmel_nand: Too many error.
atmel_nand atmel_nand: Too many error.
atmel_nand atmel_nand: Too many error.
UBI error: ubi_io_read: error -1 while reading 64 bytes from PEB 128:0, read 0 byte
slab error in kmem_cache_destroy(): cache `ubi_scan_leb_slab': Can't free all objects
Backtrace:
[<c003b898>] (dump_backtrace+0x0/0x104) from [<c0352cf0>] (dump_stack+0x18/0x1c)
r6:c7875000 r5:c7951ce0 r4:c78a30c0
[<c0352cd8>] (dump_stack+0x0/0x1c) from [<c00a6150>] (kmem_cache_destroy+0x90/0xf8)
[<c00a60c0>] (kmem_cache_destroy+0x0/0xf8) from [<c02264dc>] (ubi_scan+0x6b8/0x9ac)
r4:00000080
[<c0225e24>] (ubi_scan+0x0/0x9ac) from [<c021e4b4>] (ubi_attach_mtd_dev+0x41c/0xa08)

尝试了多次均是一样的情况。第一次烧写时，可以跑到最后的地方

VFS: Mounted root (ubifs filesystem) on device 0:12.
Freeing init memory: 136K
Failed to execute /init.  Attempting defaults...
Kernel panic - not syncing: Attempted to kill init!
Backtrace:
[<c003b898>] (dump_backtrace+0x0/0x104) from [<c0352cf0>] (dump_stack+0x18/0x1c)
r6:c781a00c r5:00000004 r4:c0489878
[<c0352cd8>] (dump_stack+0x0/0x1c) from [<c0352e70>] (panic+0x5c/0x188)
[<c0352e14>] (panic+0x0/0x188) from [<c004b8b4>] (do_exit+0x98/0x644)
r3:c046c394 r2:c7819cfc r1:c781be5c r0:c0403e44
r7:c781a00c
[<c004b81c>] (do_exit+0x0/0x644) from [<c004bef8>] (do_group_exit+0x5c/0xc4)
r7:c781cd60
[<c004be9c>] (do_group_exit+0x0/0xc4) from [<c0055eac>] (get_signal_to_deliver+0x30c/0x348)
r4:00000004
[<c0055ba0>] (get_signal_to_deliver+0x0/0x348) from [<c003ac00>] (do_signal+0xc8/0x5a0)
[<c003ab38>] (do_signal+0x0/0x5a0) from [<c003b3f8>] (do_notify_resume+0x20/0x74)
[<c003b3d8>] (do_notify_resume+0x0/0x74) from [<c0038b94>] (work_pending+0x24/0x28)
r4:00000000

但是重启后，就在ubi最开始探测的时候就已经出问题了。网上各种放狗搜，各种说法都有。

于是，各种排除。怀疑的方向主要是这样的，为什么第一次是好好的，重启后不行了呢？是因为第一次运行修改了flash的数据？？破坏了文件系统结构？如果是这样，谁修改的？是内核？？内核做什么了？？？在什么地方修改的？？？其实，这些疑问排查起来非常困难。

于是，在想，是不是ECC校验出的问题。但是转念一想，uboot 就是配置的4BIT的pmecc，kernel也是一样。亲手修改的代码啊！！！！

想了一上午，没头绪。还是看看代码吧，果不然，有个变量写错了。

}
if (nand_chip->ecc.mode == NAND_ECC_HW) {
/* ECC is calculated for the whole page (1 step) */
nand_chip->ecc.size = mtd->writesize;
//modified by zhaozx
/* set ECC page size and oob layout */
switch (mtd->writesize) {
case 2048:
nand_chip->ecc.bytes = 28;
nand_chip->ecc.steps = 1;
nand_chip->ecc.layout = &pmecc_oobinfo_2048;
host->mm = GF_DIMENSION_13;
host->nn = (1 << host->mm) - 1;
/* 4-bits correction */
host->tt = 4;
host->sector_size = 512;
host->sector_number = mtd->writesize /
host->sector_size;
host->ecc_bytes_per_sector = 7;///BUG here.这个地方官方默认是2bit校验是4。现在用4bit ecc应该是7，悲剧！！！！
host->alpha_to = pmecc_get_alpha_to(host);
host->index_of = pmecc_get_index_of(host);
break;

结论：

真相只有一个，就是在代码里！！！！

从打印入手：

关键的打印信息[UBI error: ubi_io_read: error -1 while reading 64 bytes from PEB 128:0, read 0 byte]

这句话直面的意思就是，从flash读数据，结果一个字节没读出来？？？？为什么，直观的感觉就是ecc校验出问题了。

因为用到的pmecc，校验码是硬件写入到flash的。如果此时对pmecc的寄存器的配置，与写数据时的配置不一样，那么必然会出问题。

所以，推测，问题应该出在ecc的校验！！也就是上面说的，发现代码的书写有个bug！！！