Hello, Le 11/06/2026 à 14:15, Jeff King a écrit :
Package: libccid Version: 1.8.1-1 Severity: normalAfter upgrading libccid (and restarting pcscd), pcsc_scan no longer finds my Yubikey. Bisection in the upstream git repository points to a1ccc12 (Correctly close the slots of a multi-slots reader, 2026-06-03). Building libccid with that commit and then running "pcscd --foreground --debug" produces these log entries: 00000029 ../src/ifdhandler.c:91:CreateChannelByNameOrChannel() Lun: 10000, device: usb:1050/0406:libudev:1:/dev/bus/usb/001/002 00000005 ../src/utils.c:77:GetNewReaderIndex() Lun: 10000 is already used 00000004 ../src/readerfactory.c:1144:RFInitializeReader() Open Port 0x200001 Failed (usb:1050/0406:libudev:1:/dev/bus/usb/001/002) 00000004 ../src/readerfactory.c:371:RFAddReader() Yubico YubiKey FIDO+CCID init failed. 00000006 ../src/readerfactory.c:627:RFRemoveReader() UnrefReader() count was: 1 00000003 ../src/readerfactory.c:1157:RFUnInitializeReader() Attempting shutdown of Yubico YubiKey FIDO+CCID 01 00. 00000007 ../src/hotplug_libudev.c:527:HPAddDevice() Failed adding USB device: Yubico YubiKey FIDO+CCID whereas building with the parent of that commit, the "Lun: 10000 is already used" error does not occur (and then it actually opens the device). The "CreateChannelByNameOrChannel(): Lun: 10000" line appears twice, even in the working case. So I guess we open it twice while probing for the correct driver, but the change in a1ccc12 means we do not correctly close the channel, and the second attempt then fails. I think we never hit the ReleaseReaderIndex() call in CloseUSB(), and the one in FreeChannel() is now gone. Doing this: diff --git a/src/ifdhandler.c b/src/ifdhandler.c index 63b7cd7..03a03c1 100644 --- a/src/ifdhandler.c +++ b/src/ifdhandler.c @@ -203,6 +203,7 @@ error: if (return_value != IFD_SUCCESS) { /* release the allocated resources */ + ReleaseReaderIndex(ccid_reader->reader_index); FreeChannel(ccid_reader); }fixes it for me, but I don't know the code well enough to say if that's a sane fix. It looks like FreeChannel() calls ClosePort(), which is #define'd to CloseUSB(), which is where the new release code was added. But we never hit that code, because ccid_reader->device->dev_handle is NULL and we return early. So maybe it needs to try harder to release the index in this case?
You patch is mostly correct. I fixed the bug in https://github.com/LudovicRousseau/CCID/commit/3a3373da81174ba439e7c3bca2b3678aa4de7865 I will release a new version soon. Thanks -- Dr. Ludovic Rousseau
