BBB 烧写 EMMC 不能从 EMMC 启动

user4725089

测试的板子，使用 BBB。制作 SD 卡，启动 linux，MLO uboot 内核使用的是自己的。文件系统使用的是 **AM335x Beaglebone Black eMMC烧写记录** 提供的 FlasherFS.tar.gz 。

在 linux 启动之后，执行 ./debrick.sh (已经将 DRIVE 指定为 "/dev/mmcblk1" 了)。成功创建了两个分区，mmcblk1p1 mmcblk1p2 ，且已经将 ML0 u-boot.img uExt.t uImage 解压到 /dev/mmcblk1p1 分区。文件系统解压到 /dev/ mmcblk1p2 .

可是，断电，拔掉 SD 卡，重新上电，设备不能从 EMMC 启动。一直打印 "CCCCCCCC"

IT 的朋友，求知识，求指导

8 年多前

0 user4725089 8 年多前

Expert 1027 points

帖子不能沉。

EMMC 被创建为两个分区了，第一个分区是 fat 格式，从 SD 启动 linux 确认查看了 ,MLO、u-boot.img、uExt.txt uImage, 确实已经写入到 mmcblk1p1 了。板子是 BBB 的，就是不能启动

0 user4725089 8 年多前回复 user4725089

Expert 1027 points

EMMC 使用的是 4GB 的大小，接在 MMC1 上面，型号是 emmc_ps8210_v451_80s_153b_wt.pdf。查阅了文档，EMMC 的 version 是 4.51。所以，设备接在 MMC1 接口上，应该是没有问题的。

0 user4725089 8 年多前

Expert 1027 points

我在 BBB 的公版上，验证没有通过。文件系统和脚本，是官方提供的，内核是自己的，但是是支持EXT3/4 等格式的。从 /media/mmcblk1p1 查看可以得知，MLO、u-boot.img、uExt.txt 是正确写入到 fat 分区里面了。

0 Jian Zhou 8 年多前回复 user4725089

TI__Mastermind 45260 points

烧写的log有没有？

0 user4725089 8 年多前回复 Jian Zhou

Expert 1027 points

./debrick.sh
****************************************************
****************************************************

Sitara Example Flashing Script - 02/11/2014

1+0 records in
1+0 records out
SIZE = 3825205248
CYLINDERS = 465
No partition found. Continuing.
Partitioning the eMMC...
Checking that no-one is using this disk right now ...
[ 4320.991424] mmcblk1: unknown partition table
OK

Disk /dev/mmcblk1: 465 cylinders, 255 heads, 63 sectors/track

sfdisk: ERROR: sector 0 does not have an msdos signature
sfdisk: /dev/mmcblk1: unrecognized partition table type

Old situation:
sfdisk: No partitions found

New situation:
Units = cylinders of 8225280 bytes, blocks of 1024 bytes, counting from 0

Device Boot Start End #cyls #blocks Id System
/dev/mmcblk1p1 * 0+ 89 90- 722893+ c W95 FAT32 (LBA)
/dev/mmcblk1p2 91 464 374 3004155 83 Linux
/dev/mmcblk1p3 0 - 0 0 0 Empty
/dev/mmcblk1p4 0 - 0 0 0 Empty
Successfully wrote the new partition table

Re-reading the partition table ...
[ 4321.593139] mmcblk1: p1 p2

If you created or changed a DOS partition, /dev/foo7, say, then use dd(1)
to zero the first 512 bytes: dd if=/dev/zero of=/dev/foo7 bs=512 count=1
(See fdisk(8).)
[ 4323.968322] kjournald starting. Commit interval 5 seconds
[ 4323.976013] EXT3-fs (mmcblk1p2): using internal journal
[ 4323.981536] EXT3-fs (mmcblk1p2): mounted filesystem with ordered data mode
Found mounted partition(s) on /dev/mmcblk1: /media/mmcblk1p1 /media/mmcblk1p2
4k erase on /dev/mmcblk1p1
1+0 records in
1+0 records out
4k erase on /dev/mmcblk1p2
1+0 records in
1+0 records out
1+0 records in
1+0 records out
Formatting the eMMC into 2 partitions...
mkfs.vfat 2.11 (12 Mar 2005)
mke2fs 1.42.1 (17-Feb-2012)
Discarding device blocks: done
Filesystem label=rootfs
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
Stride=0 blocks, Stripe width=0 blocks
188048 inodes, 751038 blocks
37551 blocks (5.00%) reserved for the super user
First data block=0
Maximum filesystem blocks=771751936
23 block groups
32768 blocks per group, 32768 fragments per group
8176 inodes per group
Superblock backups stored on blocks:
32768, 98304, 163840, 229376, 294912

Allocating group tables: done
Writing inode tables: done
Creating journal (16384 blocks): done
Writing superblocks and filesystem accounting information: done

Formatting done.
[ 4344.397766] kjournald starting. Commit interval 5 seconds
[ 4344.405395] EXT3-fs (mmcblk1p2): using internal journal
[ 4344.410919] EXT3-fs (mmcblk1p2): mounted filesystem with ordered data mode
Copying Files...
real 0m 0.63s
user 0m 0.40s
sys 0m 0.14s
Boot partition done.
real 0m 1.07s
user 0m 0.66s
sys 0m 0.32s
RootFS partition done.
It took 31 seconds to complete this task...

********************************************
Sitara Example Flash Script is complete.

0 user4725089 8 年多前回复 user4725089

Expert 1027 points

我将 boot_partition.tar.gz rootfs_partition.tar.gz debrick.sh 放在 /home/root 目录下面

因为是SD 卡启动的，所以对 debrick.sh DRIVER 设置为 /dev/mmcblk1

0 user4725089 8 年多前回复 user4725089

Expert 1027 points

在 SD 卡启动的 MLO 和 u-boot.img 和 EMMC 启动的 MLO、 u-boot.img 是一样的，没错吧

0 user4725089 8 年多前回复 Jian Zhou

Expert 1027 points

Jian zhou，快点出来，帮忙解决问题吧。这个问题，卡了我快一周了

0 Jian Zhou 8 年多前回复 user4725089

TI__Mastermind 45260 points

SD卡启动试过没有？你的MLO确定是可以正常启动的么？

0 user4725089 8 年多前回复 Jian Zhou

Expert 1027 points

SD卡可以启动，MLO 没有问题。 SD 卡启动了，通过 SD 卡启动了 linux（文件系统你们的 FlasherFS.tar.gz），系统起来之后，手动运行了 debrick.sh 脚本。emmc 设备 fat 分区的 MLO 和 SD 启动的 MLO，是同一份 MLO。

0 Jian Zhou 8 年多前回复 user4725089

TI__Mastermind 45260 points

那就很奇怪了，我在BBB板上验证很多次，至少MLO引导是没有问题的。

sysboot没有问题吧?

0 user4725089 8 年多前回复 Jian Zhou

Expert 1027 points

echo "****************************************************"

echo ""

echo "Sitara Example Flashing Script - 02/11/2014"

echo ""

STARTTIME=$(date +%s)

##---------Start of variables---------------------##

## Set Server IP here

#SERVER_IP="192.168.100.1"

## Names of the images to grab from TFTP server

BOOT_PARTITION="boot_partition.tar.gz"

## Rename rootfs as needed depending on use of tar or img

ROOTFS_PARTITION="rootfs_partition.tar.gz"

## ROOTFS_PARTITION="rootfs_partition.img.gz"

## Declare eMMC device name here

DRIVE="/dev/mmcblk1"

##----------End of variables-----------------------##

## TFTP files from host. Edit the files and host IP address for your application.

## We are grabbing two files, one an archive with files to populate a FAT partion,

## which we will create. Another for a filesystem image to 'dd' onto an unmounted partition.

## Using a compressed tarball can be easier to implement, however, with a large file system

## with a lot of small files, we recommend a 'dd' image of the partition to speed up writes.

## echo "Getting files from server: ${SERVER_IP}"

## time tftp -b 4096 -g -r ${BOOT_PARTITION} ${SERVER_IP} &

## boot_pid=$!

## time tftp -b 4096 -g -r ${ROOTFS_PARTITION} ${SERVER_IP} &

## rootfs_pid=$!

## Kill any partition info that might be there

dd if=/dev/zero of=$DRIVE bs=4k count=1

sync

## Figure out how big the eMMC is in bytes

SIZE=`fdisk -l $DRIVE | grep Disk | awk '{print $5}'`

echo "SIZE = $SIZE"

## Translate size into segments, which traditional tools call Cylinders. eMMC is not a spinning disk.

## We are basically ignoring what FDISK and SFDISK are reporting because it is not really accurate.

## we are translating this info to something that makes more sense for eMMC.

CYLINDERS=`echo $SIZE/255/63/512 | bc`

echo "CYLINDERS = $CYLINDERS"

## Check to see if the eMMC partitions have automatically loaded from the old MBR.

## This might have occured during the boot process if the kernel detected a filesystem

## before we killed the MBR. We will need to unmount and kill them by writing 4k zeros to the

## partitions that were found.

check_mounted(){

is_mounted=$(grep ${DRIVE}p /proc/mounts | awk '{print $2}')

if grep -q ${DRIVE}p /proc/mounts; then

echo "Found mounted partition(s) on " ${DRIVE}": " $is_mounted

umount $is_mounted

counter=1

for i in $is_mounted; do \

echo "4k erase on ${DRIVE}p${counter}";

dd if=/dev/zero of=${DRIVE}p${counter} bs=4k count=1;

counter=$((counter+1));

done

else

echo "No partition found. Continuing."

}

check_mounted;

## Partitioning the eMMC using information gathered.

## Here is where you can add/remove partitions.

## We are building 2 partitions:

## 1. FAT, size = 9 cylinders * 255 heads * 63 sectors * 512 bytes/sec = ~70MB

## 2. EXT3, size = 223 ($CYLINDERS-[9 for fat]) cylinders * 255 heads * 63 sectors * 512 bytes/sec = ~1l.7GB

## You will need to change the lines ",9,0c0C,*", "10,,,-" to suit your needs. Adding is similar,

## but you will need to be aware of partition sizes and boundaries. Use the man page for sfdisk.

echo "Partitioning the eMMC..."

sfdisk -D -H 255 -S 63 -C $CYLINDERS $DRIVE << EOF

,90,0x0C,*

91,,,-

EOF

## This sleep is necessary as there is a service which attempts

## to automount any filesystems that it finds as soon as sfdisk

## finishes partitioning. We sleep to let it run. May need to

## be lengthened if you have more partitions.

sleep 2

## Check here if there has been a partition that automounted.

## This will eliminate the old partition that gets

## automatically found after the sfdisk command. It ONLY

## gets found if there was a previous file system on the same

## partition boundary. Happens when running this script more than once.

## To fix, we just unmount and write some zeros to it.

check_mounted;

## Clean up the dos (FAT) partition as recommended by SFDISK

dd if=/dev/zero of=${DRIVE}p1 bs=512 count=1

## Make sure posted writes are cleaned up

sync

## Format the eMMC into 2 partitions

echo "Formatting the eMMC into 2 partitions..."

## Format the boot partition to fat32

mkfs.vfat -F 32 -n "boot" ${DRIVE}p1

## Format the rootfs to ext3 (or ext4, etc.) if using a tar file.

## We DO NOT need to format this partition if we are 'dd'ing an image

## Comment out this line if using 'dd' of an image.

mkfs.ext3 -L "rootfs" ${DRIVE}p2

## Make sure posted writes are cleaned up

sync

echo "Formatting done."

## Make temp directories for mountpoints

mkdir tmp_boot

## Comment this line out if using 'dd' of an image. It is not needed.

mkdir tmp_rootfs

## Mount partitions for tarball extraction. NOT for 'dd'.

mount -t vfat ${DRIVE}p1 tmp_boot

## If 'dd'ing the rootfs, there is no need to mount it. Comment out the below.

mount -t ext3 ${DRIVE}p2 tmp_rootfs

## Wait for boot to finish tftp

wait $boot_pid

echo "Copying Files..."

time tar -xf ${BOOT_PARTITION} -C tmp_boot

sync

umount ${DRIVE}p1

##rm ${BOOT_PARTITION}

echo "Boot partition done."

## Wait for rootfs to finish tftp

wait $rootfs_pid

## If using a tar archive, untar it with the below.

## If using 'dd' of an img, comment these lines out and use the below.

time tar -xf ${ROOTFS_PARTITION} -C tmp_rootfs

sync

umount ${DRIVE}p2

##rm ${ROOTFS_PARTITION}

echo "RootFS partition done."

## If using 'dd' of an img, uncomment these lines.

## If using a tar archive, comment out these lines and use the above.

## time gunzip -c ${ROOTFS_PARTITION} | dd of=${DRIVE}p2 bs=4k

## sync

## echo "Rootfs partition done."

## The block below is only necessary if using 'dd'.

## Force check the filesystem for consistency and fix errors if any.

## Resize partition to the length specified by the MBR.

## /sbin/e2fsck -fy ${DRIVE}p2

## /sbin/resize2fs ${DRIVE}p2

ENDTIME=$(date +%s)

echo "It took $(($ENDTIME - $STARTTIME)) seconds to complete this task..."

## Reboot

echo ""

echo "********************************************"

echo "Sitara Example Flash Script is complete."

echo ""

这个脚本，是你们的，几乎没怎么修改，

0 user4725089 8 年多前回复 Jian Zhou

Expert 1027 points

sysboot 的话， BBB 的板子进行进行测试，默认的 11100 ，MMC1-> MMC0 -》 UART0 -》 USB0。

咨询下，你可以将你们在 BBB 测试的，所有文件，重新打包给我不。我现在不知道，感觉都是按照官方的流程走的，不知道是哪里出了问题。

0 user4725089 8 年多前回复 user4725089

Expert 1027 points

能不能帮忙我排除一下问题，我这边实在定位不到问题了。

能不能麻烦你们，用我刚发的脚本，做测试。确认下，是不是脚本的问题，还是其他的问题

0 Jian Zhou 8 年多前回复 user4725089

TI__Mastermind 45260 points

你手上有几个BBB的板子？有没换个板子试试。

如果硬件正常，烧进去的MLO至少能跑起来。

0 user4725089 8 年多前回复 Jian Zhou

Expert 1027 points

我有两个 BBB 的板子，但是都不能起来。

如果，你那边可以的话，能不能麻烦你，用我前面的脚本，去跑，如果一切正常，顺利从 EMMC 启动的话，说明脚本是没有问题的。

现在我这边，定位不到问题。

0 Jian Zhou 8 年多前回复 user4725089

TI__Mastermind 45260 points

你用的是SDK06的MLO么？如果不是，用附件的试下。

MLO-am335x-evm

处理器

处理器论坛

BBB 烧写 EMMC 不能从 EMMC 启动