OpenPGP: fixed state tracking after erasing card

[frankmorgner]

fixes #3018

Checklist

• Documentation is added or updated
• New files have a LGPL 2.1 license statement
• PKCS#11 module is tested
• Windows minidriver is tested
• macOS tokend is tested

[Jakuje]

If the pgp_new_blob() fails, the memory allocated by sc_file_new() is leaked (yeah, it was in the original code too but as I noticed it, it would be good to fix it).

[frankmorgner]

thanks for the review. I'll look at the suggestions, when @arrowd confirms that it actually fixes the problem

[arrowd]

I will not be able to test this during this week, sorry, but I will certainly check this out later.

If I'm reading it right, this change would allow me to just do sc_disconnect_card + sc_connect_card to refresh the state?

[frankmorgner]

If I'm reading it right, this change would allow me to just do sc_disconnect_card + sc_connect_card to refresh the state?

No. If this works as expected, then you will get refreshed objects right after erasing the card without the need of reconnecting.

[arrowd]

The problem is not only with erasing.

The first post of #3018 shows a code that generates a keypair twice in a row, but it seemingly doesn't change until the reinsertion (physical or emulated with my hacks).

[frankmorgner]

please check if it works in case of erasing the card. If so, one case is fixed and you know where to poke.

[arrowd]

Sorry for taking this too long. The proposed change indeed improves things for both opensc-explorer and my daemon program. Erasing the card causes OPENPGP_ACCOUNT field to be reset immediately without the need to reinsert the token or to programmatically reset the reader via hacks.

The key information isn't updated, though.

[frankmorgner]

If I understand correctly, you are NOT referring to opensc-explorer with respect to the key inforamtion, since this isn't used anywhere here. I assume that opensc-explorer works as expected.

If you're talking about the card's profile which is readable via PKCS#11, we will only refresh that if modifications were executed via PKCS#11. If some concurrent process modifies the card, you need to C_Finalize and C_Initialize again to refresh the token information. Modification of the OpenPGP card via PKCS#11 is not completely implemented; in particular, erasing the card is currently not available:

OpenSC/src/pkcs15init/pkcs15-openpgp.c

Lines 45 to 48 in de5942f

	static int openpgp_erase(struct sc_profile *profile, sc_pkcs15_card_t *p15card)
	{
	return SC_ERROR_NOT_SUPPORTED;
	}

[dengert]

Modification of the OpenPGP card via PKCS#11 is not completely implemented; in particular, erasing the card is currently not available:

It might be easy to add. OpenPGP specs V3.4.1 (also in v.2.0 ) defines two optional commands
7.2.16 TERMINATE DF must be run first.
7.2.17 ACTIVATE FILE
"This optional command (announced in Life Cycle Status indicator in Historical bytes) can
be used to reset all values (DOs, Keys, PWs etc.) to their default values (after production/
initialisation)."

openpgp-tool.c uses SC_CARDCTL_ERASE_CARD to call https://github.com/OpenSC/OpenSC/blob/master/src/libopensc/card-openpgp.c#L3480-L3555

Accept for the pin code this comes down to the TERMINATE DF and ACTIVATE FILE commands.

@arrowd you could try these on your cards" opensc-tool -s "00 E6 00 00" -s "00 44 00 00"
but the could leave the card unusable

[arrowd]

If I understand correctly, you are NOT referring to opensc-explorer with respect to the key inforamtion, since this isn't used anywhere here. I assume that opensc-explorer works as expected.

Yes, the problem with opensc-explorer was

OpenSC [3F00]> cat 005e
00000000: 61 73 64 61 73 64 00 asdasd.
OpenSC [3F00]> erase
OpenSC [3F00]> cat 005e
00000000: 61 73 64 61 73 64 00 asdasd.

It now works correctly - nothing gets printed after erase is done.

If you're talking about the card's profile which is readable via PKCS#11, we will only refresh that if modifications were executed via PKCS#11. If some concurrent process modifies the card, you need to C_Finalize and C_Initialize again to refresh the token information.

There are no concurrent processes in my case. I just call C_GenerateKeyPair twice and the key doesn't change after the second call without reinserting/resetting hack. See #3018 (comment) after the "A related example with PKCS11 API is following" sentence.

C_Finalize was already suggested before, but it isn't a solution for me: #3018 (comment)

Modification of the OpenPGP card via PKCS#11 is not completely implemented; in particular, erasing the card is currently not available

Indeed, in my daemon I call sc_card_ctl(m_slot->p11card->card, SC_CARDCTL_ERASE_CARD, NULL) to reset the token. I tried C_InitToken but it didn't work.

you could try these on your cards" opensc-tool -s "00 E6 00 00" -s "00 44 00 00"
but the could leave the card unusable

What does 00 44 00 00 do in this case?

[dengert]

What does 00 44 00 00 do in this case?

7.2.17 ACTIVATE FILE
This optional command (announced in Life Cycle Status indicator in Historical bytes) can
be used to reset all values (DOs, Keys, PWs etc.) to their default values (after production/
initialisation). The command has effect only, if the life cycle status is in initialisation state
(indicator byte set to 03). If the command is used in operational state, it will end with status
bytes 9000, but nothing in the application will be changed (dummy command). The command
can be used directly after TERMINATE DF without resetting the card.
Command:
CLA 00 / 0C
INS 44
P1 00
P2 00
Lc Empty
Data
field
Empty
Le Empty
Response:
Data field Empty
SW1-SW2 9000 or specific status bytes
Example:
CLA INS P1 P2
00 44 00 00
SW1SW2
9000

[dengert]

00 44 00 00 "reset all values (DOs, Keys, PWs etc.) to their default values (after production/initialisation)" It is part of card-openpgp.c pgp_erase_card

If you then run opensc-explorer erase it may delete all the DOs which were just reset! I would suggest only use opensc-explorer to examine the card, not to change anything on it.

Modify pkcs15init/pkcs15-openpgp.c to call sc_card_ctl(m_slot->p11card->card, SC_CARDCTL_ERASE_CARD, NULL) (with changes as needed).

Modify libopensc/card-openpgp.c to make sure populate_blobs_to_mf is called at the end of pgp_erase_card

[frankmorgner]

I fixed the memory leaks and found some other minor fix. I'd like to leave this fix as is, because I doubt that we will get more insights without debugging.

[Jakuje]

lgtm. Do we want this in 0.25.0?

[frankmorgner]

It is not essential, so we can postpone this to the next release