Hello,

A lot of thanks for your work.

See my observations /infra/. I hope it can help you.

Best regards

Le 13/03/2025 à 19:52, Alexandre Esse a écrit :
Hello Frédéric,

Great, I didn't see you already did some integration developments on this board, I just joined the mailing list and didn't look extensively into the history. For now I only pushed the chopstx part but indeed, gnuk itself should also be updated.

I also have the STM32F103CBT6 version of the board. (marked as v1.1 on the PCB: not sure what it means: I opened an Issue on github and send an email to WeAct support to get some info: https://github.com/WeActStudio/BluePill-Plus/issues/19)
You can find the difference between V1.0 et V1.1 here : https://github.com/WeActStudio/WeActStudio.BluePill-Plus-CH32/tree/master/HDK It's for CH32 processor but the design of the boards are all the same.

I have been testing both on the 1.2.20 branch and 2.2. But I guess I will stay on v2.2 for the rest of my tests.

Here are the remaining tweaks I did on v2.2 (de9652726b1ce52b21e939c6989dda0268b5c640)of gnuk to make it work:

diff --git a/src/configure b/src/configure
index 1188a72..4ff7d1a 100755
--- a/src/configure
+++ b/src/configure
@@ -130,6 +130,7 @@ Configuration:
                           ST_NUCLEO_F103
                           NITROKEY_START
                           BLUE_PILL
+                          BLUE_PILL_PLUS
                           STM8S_DISCOVERY
                           CQ_STARM
                           STM32_PRIMER2
@@ -164,7 +165,7 @@ MEMORY_SIZE=20

 # Settings for TARGET
 case $target in
-BLUE_PILL|STM8S_DISCOVERY)
+BLUE_PILL|BLUE_PILL_PLUS|STM8S_DISCOVERY)
   # It's 64KB version of STM32F103, but actually has 128KB
   flash_override="-DSTM32F103_OVERRIDE_FLASH_SIZE_KB=128"
   ;;

There is STM32F103C6T6 (64K)  and CBT6 (128K), so i think your add relative to : # setting to target, is not necessary for CBT6 processor which have 128K memory. That's why i suggested to have multiple definition board to take care of different processor (arm and riscv) and different amount of memory
I haven't been testing "on-the-token" key generation. I only transferred to it from a host PC. I always had KDF-DO activated. The main issue I get is when I try to "reset" the token, it simply doesn't work but I haven't been investigating it. Also some PIN management's actions seem shaky (but there is a possibility that this is due to my lack of experience on gnuk tokens too). So for now, I tested the key with this kind of secret keys I get with 'gpg --list-secret-keys':
sec>  ed25519 2025-03-13 [SC]
      XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
      Card serial no. = FFFF 00000000
uid           [ultimate] Tmp Tmp <[email protected]>
ssb>  cv25519 2025-03-13 [E]
ssb>  ed25519 2025-03-13 [A]
I managed to sign, decrypt data and authenticate through ssh sessions with it. PB2 as LED is working and PA0 as ACK button is also working fine, I haven't seen any issue for these use-case in a week.

About ack button, it is on PA0 and PA0 is used for it's ADC function and entropy generation, so i think you have to modifiy /gnuk/chopstx/contrib/adc-stm32ff103.c in order to use an other pin for ADC than PA0

Entropy generation graph is explain on Niibe site here : https://www.gniibe.org/memo/development/gnuk/rng/neug.html

You can see that PA0 and PA1 are used for their ADC. It's also indicate for example in board-fst-01-00.h file

 * PA0  - input with pull-up.  AN0
 * PA1  - input with pull-up.  AN1

For Blue-Pill-Plus, button is connect to 3.3V (https://github.com/WeActStudio/BluePill-Plus/blob/master/HDK/BluePillPlus_V10_SchDoc.pdf) so i think you configure PA0 as input pull-down. I think all this change can interfere with good entropy generation. I see different cases of board in /gnuk/chopstx/contrib/adc-stm32ff103.c which seem to use different ADC pins but i don't really understand how. It's for that, i didn't take care of ackbutton to day.

---------------------------------------------------------------

In your board file, you indicate id board : 0x49403d56. In mine, i indicate 0x1ba01477. I get this id with stlink program as core id. Can your tell me how you get your board id ?

Frédéric SUEL


Regards,
Alexandre



On Fri, 7 Mar 2025 at 10:00, Frédéric SUEL <[email protected]> wrote:

    Ref : Post on the gnuk list : Frédéric SUEL frederic.suel at
    free.fr <http://free.fr> Mon Feb 17 11:13:25 CET 2025

    Hi,

    Thank you for your interest for this board.

    When i asked help about this board, i proposed a file
    board-blue-pill-plus-cb.h because this board exist with 4 arm
    processor and two riscv processors. I indicated cd because there
    is STM32F103C8T6 (64k)  and CBT6 (128k). To take care of 64K
    version, you have to add code in /gnuk/src/.configure. (see my post)

    For my blue-pilll-board stm32F103CB, i have #define BOARD_ID   
    0x1ba01477 (see my post). Perhaps your board is a STM32F103C6T6
    board, i don't know why IDs are different

    For instance, i doesn't take care of ackbutton because there is a
    supplementary problem as PA0 is used for it's ADC and for entropy
    generation. I think you have to modify
    /gnuk/chopstx/contrib/adc-stm32ff103.c but i doesn't know how.

    ----

    So i just created /gnuk/board/board-blue-pill-plus-cb.h, modified
    /gnuk/src/.configure to add definition of BLUE-PILL-PLLUS-CB
    and/**/chopstx/mcu/sys-stm31f103.h for my first tests and doesn't
    take care of ackbutton on PA0

    I get :

    -- works fine with curve25519 : generation on the board and import
    on the board.

    -- impossibility to generate secp256k1 on the board even with
    KDF-DO activate as Niibe suggested (msg : used conditions not
    satisfied)

    -- impossibility to generate X448 on the board even with Niibe
    patch (msg : board error)

    -- impossibility to import X448 on the board. It seems to work but
    only encrypt key is on the board and works. The other keys are
    marked as # .  I get with gpg --list-secret-keys

    sec#  ed448/0xAA988F88C70C3DEE 2025-02-23 [SC] [expire : 2075-02-11]

    Empreinte de la clef = AA988 F88C7 0C3DE E74BE DFF48 D127D 4BA4E
    CAEB3 685B3 575E7

    uid                  [  ultime ] tmp

    ssb>  cv448/0x406CC6562774BC84 2025-02-23 [E] [expire : 2075-02-11]

    ssb#  ed448/0x02BB1F8E7A2B268A 2025-02-23 [A] [expire : 2075-02-11]

    ----

    Can you precise what's work with your board ?

    Best regards



    Le 06/03/2025 à 23:47, Alexandre Esse a écrit :
    Hello,

    Here is a short message to notify the mailing list that I
    proposed a merge request on chopstx:
    https://salsa.debian.org/gnuk-team/chopstx/chopstx/-/merge_requests/1

    This is the first step to add gnuk support for Blue Pill Plus
    boards <https://github.com/WeActStudio/BluePill-Plus/>.

    Not sure if this is the way to contribute: looking forward to
    your feedback.

    Regards,
    Alexandre

    _______________________________________________
    Gnuk-users mailing list
    [email protected]
    https://lists.gnupg.org/mailman/listinfo/gnuk-users
_______________________________________________
Gnuk-users mailing list
[email protected]
https://lists.gnupg.org/mailman/listinfo/gnuk-users

Reply via email to