Post

Recovering the BIOS password from a Panasonic CF-U1 mk2 (AMI Aptio UEFI)

Posted
By en4rab
8 min read

This was originally posted as a gist here: https://gist.github.com/en4rab/550880c099b5194fbbf3039e3c8ab6fd

A mess of my own making

While messing with a CF-U1 handheld PC that I bought off ebay I managed to mess up the BIOS and it seems it reverted to previous settings which included an unknown BIOS password, it would however still boot into windows. Since I could still boot windows I was able to dump the bios flash using AFUWINGUI.EXE the version I used was 3.09.03.1462 which is available here:
https://ami.com/en/?Aptio_4_AMI_Firmware_Update_Utility.zip

There may be a more appropriate version to use as this seemed to have trouble checking the bios version when flashing but did work if you selected “Do Not Check ROM ID” but flashing isnt needed to get the password.

Dumping the flash

alt text
Run AFUWINGUI.EXE and at the bottom of the “Information” tab click the save button to make a backup of your bios, the default name is afuwin.rom Now open this saved image with UEFITool_NE available here:
https://github.com/LongSoft/UEFITool/releases

I used UEFITool_NE_A51_win32.zip later versions should work fine. The new engine (NE) verson seems to deal with AMI’s odd nvram format better.

alt text

Expand the first EfiFirmwareFilesystemGuid » NVRAM dropdown tree and look for the GUID
C811FA38-42C8-4579-A9BB-60E94EDDFB34 (AMITSESetup)
with subtype Data there will be others with subtype Link which are older no longer valid entrys because of the odd way AMI nvram works, if you find one of these right click on it and select “Go to data” and it will take you to the actual data entry.
Now right click and select “Body hex view” and you should see something like:

1
2
3
4
5
6
7
8
9

0000  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
0010  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
0020  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
0030  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
0040  7B 13 94 A6 07 3A 29 CD D2 60 1A F4 5C 87 ED 1A  {.”¦.:)ÍÒ`.ô\‡í.
0050  07 AE AE 41 DC D4 0A 68 AB FB FA 0E 55 A2 B0 35  .®®AÜÔ.h«ûú.U¢°5
0060  0B C9 66 5C C1 EF 1C 83 77 16 D2 A9 2D 3D 88 D0  .Éf\Áï.ƒw.Ò©-=ˆÐ
0070  E3 63 3E F7 99 8A F4 1D 4F B1 AA 44 05 D8 60 6B  ãc>÷™Šô.O±ªD.Ø`k
0080  01

In this the bytes from 0x00 to 0x3F are the currently unset user password, 0x40 to 0x7F are the obfuscated administrator password and 0x80 is the quiet boot flag.

1337 encryption

The password is obfuscated using super secure xor

VOID PasswordEncode( CHAR16 *Password, UINTN MaxSize)
{
    UINTN	ii;
    unsigned int key = 0x935b;

#if SETUP_PASSWORD_NON_CASE_SENSITIVE
    for ( ii = 0; ii < MaxSize; ii++ )
        Password[ii] = ((Password[ii]>=L'a')&&(Password[ii]<=L'z'))?(Password[ii]+L'A'-L'a'):Password[ii];
#endif

    // Encode the password..
    for ( ii = 1; ii <= MaxSize/2; ii++ )
        Password[ii-1] = (CHAR16)(Password[ii-1] ^ (key*ii));
}

So Xoring the above encoded password:

1
2

7B 13 94 A6 07 3A 29 CD D2 60 1A F4 5C 87 ED 1A 07 AE AE 41 DC D4 0A 68 AB FB FA 0E 55 A2 B0 35 
0B C9 66 5C C1 EF 1C 83 77 16 D2 A9 2D 3D 88 D0 E3 63 3E F7 99 8A F4 1D 4F B1 AA 44 05 D8 60 6B

with

1
2

5B 93 B6 26 11 BA 6C 4D C7 E0 22 74 7D 07 D8 9A 33 2E 8E C1 E9 54 44 E8 9F 7B FA 0E 55 A2 B0 35 
0B C9 66 5C C1 EF 1C 83 77 16 D2 A9 2D 3D 88 D0 E3 63 3E F7 99 8A F4 1D 4F B1 AA 44 05 D8 60 6B

gives

1
2

20 80 22 80 16 80 45 80 15 80 38 80 21 80 35 80 34 80 20 80 35 80 4e 80 34 80 00 00 00 00 00 00 
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Each character of the password is stored as 2 bytes, and as x86 is wrong endian im guessing should be read as 0x8020 0x8022 I have no idea where the 0x80 comes from possibly its something to do with the EFI_SHIFT_STATE_VALID in this case the password was lower case, possibly uppercase status is encoded in this byte too I have no idea I havent tested uppercase passwords.

WTF scancodes how does this map to keys

From the unobfuscated data you can see the password is 13 characters long, im going to ignore the 0x80 bytes as i dont understand them :P and just look at the others:
20 22 16 45 15 38 21 35 34 20 35 4e 34
They appear to be some sort of scancodes, although while googleing this I found some AMI bioses seem to use ascii here so you can read it out directly as text, but not on this machine.
When this CF-U1 arrived from ebay it had a password which i sucessfully guessed as “toughbook” my second guess would have been “panasonic” since using text written on the front of the PC as a password saves writing it under the battery cover :P
Looking through the older link entrys for the AMITSESetup nvram I found what I thought was the data for this password which deobfuscating as above gave (ignoring the 0x80):

1
2

35 39 37 24 25 14 39 39 27
t  o  u  g  h  b  o  o  k

This seemed promising repeated characters have the same value and gives a bit of a key to the mapping Some googeling later about UEFI scancodes and i found this page:
https://web.archive.org/web/20170627134759/http://wiki.phoenix.com/wiki/index.php/EFI_KEY
From this it seems the value is the offset into this enum so in the toughbook example 35 translates to EfiKeyD5 a second page I found gave the mapping from EfiKey to ascii:
https://github.com/tianocore/edk2/blob/master/MdeModulePkg/Bus/Usb/UsbKbDxe/KeyBoard.c#L36

So i made up a list of byte to ascii using these, below are just 0x10 to 0x4E to cover most values but not be too stupidly long.

HexCharEFIkeyHexCharEFIkey
10zEfiKeyB130TabEfiKeyTab
11xEfiKeyB231qEfiKeyD1
12cEfiKeyB332wEfiKeyD2
13vEfiKeyB433eEfiKeyD3
14bEfiKeyB534rEfiKeyD4
15nEfiKeyB635tEfiKeyD5
16mEfiKeyB736yEfiKeyD6
17,EfiKeyB837uEfiKeyD7
18.EfiKeyB938iEfiKeyD8
19/EfiKeyB1039oEfiKeyD9
1A EfiKeyRShift3ApEfiKeyD10
1B EfiKeyUpArrow3B[EfiKeyD11
1C1EfiKeyOne3C]EfiKeyD12
1D2EfiKeyTwo3D\EfiKeyD13
1E3EfiKeyThree3E EfiKeyDel
1F EfiKeyCapsLock3F EfiKeyEnd
20aEfiKeyC140 EfiKeyPgDn
21sEfiKeyC2417EfiKeySeven
22dEfiKeyC3428EfiKeyEight
23fEfiKeyC4439EfiKeyNine
24gEfiKeyC544`EfiKeyE0
25hEfiKeyC6451EfiKeyE1
26jEfiKeyC7462EfiKeyE2
27kEfiKeyC8473EfiKeyE3
28lEfiKeyC9484EfiKeyE4
29;EfiKeyC10495EfiKeyE5
2A'EfiKeyC114A6EfiKeyE6
2B|EfiKeyC124B7EfiKeyE7
2C4EfiKeyFour4C8EfiKeyE8
2D5EfiKeyFive4D9EfiKeyE9
2E6EfiKeySix4E0EfiKeyE10
2F+EfiKeyPlus   

So what was the password?

Using the above list and the recovered scancodes gave:

1
2

20 22 16 45 15 38 21 35 34 20 35 4e 34
a  d  m  1  n  i  s  t  r  a  t  0  r

and when i tried adm1nistrat0r it worked!
This is not complete as there are still questions about the 0x80 bytes but my guess is they encode the shift alt etc modifier keys but im back into my handheld so i’m not sure ill look further into it. This may also apply to other Aptio bioses as well as the Panasonic CF-U1, and if the machine isnt bootable you may be able to use a cheap spi adapter to dump the bios, in the case of the CF-U1 it uses an LPC flash which I don’t think you can get cheap clips and readers for and its buried in the machine so a nuisance to get to.

I just want to remove the password

Another option instread of trying to decode the password or if the password is stored as a sha1/sha256 hash is to dump the bios chip using an spi programmer. Open the spi dump and find the AMITSESetup data and just edit it with a hex editor to replace the obfuscated password with 00’s. This could then be flashed back using the spi flasher. Be aware on more modern AMI uefi’s there may be two copys of the NVRAM and the AMITSESetup needs to be edited in both copys.

Research
Research UEFI
This post is licensed under CC BY 4.0 by the author.