SG-5100: Completely wipe eMMC after M.2 installation

HG

Hi,

I have installed an M.2 drive and reinstalled the Netgate pfSense factory image which worked great, everything is up and running fine.

During the installation I also deleted mmcsd0 and after the installation I also did a "dd if=/dev/zero of=/dev/mmcsd0 bs=1M" which also succeeded (said it had written about 8GB) and rebooted. However dmesg still shows the partitions:

mmcsd0: 8GB <MMCHC M32508 5.2 SN xxxxxxxx MFG 08/2018 by 112 0x0000> at mmc0 50.0MHz/8bit/65535-block
mmcsd0boot0: 4MB partion 1 at mmcsd0
mmcsd0boot1: 4MB partion 2 at mmcsd0
mmcsd0rpmb: 4MB partion 3 at mmcsd0

And I also see them at /dev/.

What else do I have to do to remove them? gpart doesn't see them, but I must admit that I'm not exactly an expert with gpart on FreeBSD. As the kernel still sees these partitions, I'm also not sure if my dd was successful? I would have expected that it wiped everything including partition table etc.

I already read https://forum.netgate.com/topic/121270/clear-emmc-on-sg-2440-with-ssd-installed but this didn't really help me except for confirming that what I did wasn't so wrong.

stephenw10

Hmm, curious. What does geom label status show?

HG

It shows:

                  Name  Status  Components
  diskid/DISK-xxxxxxxx     N/A  mmcsd0
ufsid/xxxxxxxxxxxxxxxx     N/A  ada0s1a
           label/swap0     N/A  ada0s1b

HG

Sorry for bringing this up again... Any idea anyone?

I also did a hexdump -n 104857600 /dev/mmcsd0 and it shows that at least the first 100 MB are all zero:

0000000 0000 0000 0000 0000 0000 0000 0000 0000
*
6400000

So no partition table or something... Is there something special about these eMMCs that they always report these partitions to the kernel somehow?

stephenw10

You sometimes see a backup partition table at a later point in the drive that is not removed by erasing the primary table. Sometimes that is detected and read.
Is this actually causing a problem for you though?

Steve

HG

Thanks Steve.

@stephenw10 said in SG-5100: Completely wipe eMMC after M.2 installation:

You sometimes see a backup partition table at a later point in the drive that is not removed by erasing the primary table. Sometimes that is detected and read.

I should then find this when I do a full hexdump of /dev/mmcsd0, right? If it's still somewhere there, that then probably also means that I did something wrong with my dd...

@stephenw10 said in SG-5100: Completely wipe eMMC after M.2 installation:

Is this actually causing a problem for you though?

No, no actual problem in terms of misbehavior. Just want to make sure that I have wiped everything correctly and also want to understand the state I currently have which feels a bit "unclean".

HG

Just did a full hexdump:

[2.4.5-RELEASE][admin@...]/root: hexdump /dev/mmcsd0
0000000 0000 0000 0000 0000 0000 0000 0000 0000
*
1d2000000

The strange thing is that it returned exactly 7GB instead of 8GB.

stephenw10

Ok, try gpart list I think that will show the three partitions on there.

You should be able to remove them by index number using gpart delete -i 1 diskid/DISK-xxxxxxxx.

You don't need to though. And you'd better be sure you are removing the right thing.

Steve

HG

Already tried that as well, also when i do gpart list -a. It only shows my ada0 stuff from the SSD, but no trace of mmcsd0. Same for gpart status or gpart status -a. Really don't understand where the kernel gets this from.

gpart delete -i 1 diskid/DISK-xxxxxxxx (with my ID from geom label status of course) returns gpart: arg0 'diskid/DISK-xxxxxxxx': Invalid argument.

stephenw10

Hmm, well you might try creating a partition with gpart. I would think that will create a new table if it cannot read one.
You could then just leave that or try removing it.

Steve