Currently, if a smartcard user tries to e.g. encrypt/sign a file and enters a bad PIN, the pinentry process is cancelled and the error displayed in Kleopatra:
To be able to retry, the whole encrypt/sign process has to be started again (choosing the file, etc), which is tedious.
Pinentry should handle the retries on bad PIN instead, which would also meet the user's expectation.
Split from T6425: improve pinentry behavior and texts in smart card context
| rG GnuPG | |||
| rGfacf699862fa scd: Add askpin_inquiry for app-openpgp, not enabled for now. | |||
| rGe660a0e145dc scd: Factor out pinpad_prompt, clean-up for app-openpgp. | |||
| rGdc03307f3175 agent: Implement ASKPIN/NEXTPIN/FINISHPIN inquiry in the agent side. | |||
| rG1f039374d8a4 Simplify the call-chain to scd_getpin. | |||
| rG3add2ab5d1a0 agent: Expose scd_getpin and remove getpin_cb callback. | |||
| rG76b47528526e scd: Simplify not to use internal PIN_CB callback. | |||
| Status | Assigned | Task | ||
|---|---|---|---|---|
| Open | None | T6785 Kleopatra: Improvements related to smart cards | ||
| Open | • gniibe | T8353 Let pinentry or gnupg handle bad smartcard pins | ||
| Open | • gniibe | T8520 scd: pincache_put/pincache_get with AESWRAP with Padding (RFC-5649) |
I don't think that Kleopatra can influence that gpg/gpgsm/gpg-agent shows a second pinentry if on the first try a wrong passphrase was entered. Therefore I think the "retry" has to be implemented in GnuPG.
Independently of this, explicit canceling of the pinentry shouldn't abort the sign/encrypt operation in Kleopatra.
So, this ticket is actually two.
I have created T8433: Kleopatra: Allow easy retry of sign/encrypt files operation if password request failed or was aborted for improving Kleopatra. This leaves the problem that entering a bad smartcard PIN (or password) aborts a sign operation without repeating the PIN/password request. I think this needs to be addressed in GnuPG. -> I re-tag this ticket and request another triage.
I guess it's fine to decide not to change GnuPG. Once Kleopatra allows for an easy retry of a sign/encrypt operation a change in GnuPG will be unnecessary for users of Kleopatra.
For OpenPGP card, here is an attempt for supporting the idea:
Reading the code around, it seems for me that adding loop here means loosing simplicity of the design, though.
We need to do similar things for each card supported. And also for gpg-agent, in case of key in a file.
gpg-agent already has a loop for asking multiple times for a PIN/passphrase:
for (;pininfo->failed_tries < pininfo->max_tries; pininfo->failed_tries++)
When asked for the passphrase for a signing key stored on disk I've seen a second pinentry (with "try 2 of 3"). I'm not sure why it isn't used when asking for the PIN.
@ikloecker Thank you for pointing out how gpg-agent works.
Indeed, in gpg-agent, it uses agent_askpin with the callback to unprotect the key in a file; When the callback fails, it loops to ask a user again.
Currently, scdaemon uses assuan_inquire to let gpg-agent call agent_askpin, and in this use case, the callback is not used.
Let me consider again, if we can utilize the feature of agent_askpin callback by scdaemon.
I created a branch gniibe/t8353 for this task.
My plan is:
Here is an idea to enhance the inquiry protocol from scdaemon to gpg-agent:
Currently, it's simply:
scdaemon --NEEDPIN--> gpg-agent
<---- pin --Extended version will be:
scdaemon --NEEDPIN_BEGIN--> gpg-agent
<-- OK
-- GETPIN -->
<--- pin 0 --
-- GETPIN failure-message -->
<--- pin 1 --
...
--NEEDPIN_END-->
<-- OKgpg-agent creates a thread for pinentry interactions by NEEDPIN_BEGIN inquiry.
It interacts by following GETPIN inquiries, and finished by NEEDPIN_END inquiry.