* scd/scdaemon.c (debug_flags): Add "card".
* scd/scdaemon.h (DBG_CARD_VALUE, DBG_CARD): New.
--
Some information from parsing the card are often very helpful.
However, the card_io triggered APDU dumps are in most cases too heavy.
Thus this new debug flag.
* scd/command.c (scd_command_handler): Use gnupg_fd_t for the argument
but no INT2FD to listen. Use GNUPG_INVALID_FD.
* tpm2d/command.c (tpm2d_command_handler): Likewise.
* scd/scdaemon.c (start_connection_thread): Follow the change.
* tpm2d/tpm2daemon.c (start_connection_thread): Likewise.
* scd/scdaemon.h (scd_command_handler): Use gnupg_fd_t.
* tpm2d/tpm2daemon.h (tpm2d_command_handler): Likewise.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/scdaemon.c (main): Use gnupg_fd_t for socket, and use
assuan_sock_close for the socket allocated by assuan_sock_new.
(handle_connections): Use gnupg_fd_t for listen_fd.
Use assuan_sock_close for the socket by npth_accept.
--
GnuPG-bug-id: 5029
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/apdu.c (close_pcsc_reader): Do not decrement refcount if already
zero. Always release context if or becomes zero.
(apdu_dev_list_start): Unlock prior to close_pcsc_reader. For PC/SC
increment the count. Always release the lock.
(apdu_dev_list_finish): No more unlocking. Use close_pcsc_reader
instead of code duplication.
* scd/apdu.c (pcsc_error_string): Add an error code.
* scd/scdaemon.c (scd_kick_the_loop): Fix a diagnostic.
--
There was an obvious bug in that the pcsc.count could go below zero
and thus there was no chance to get the context release. Releasing
and recreating the context is at least under Windows important to get
rit of the PCSC_E_SERVICE_STOPPED.
Also removes a potential problem in holding the reader_table_lock
between calls to apdu_dev_list_start apdu_dev_list_finish. There is
no need for this. Instead we bump the pcsc.count.
The reader_table_lock strategy should be reviewed; we may be able to
remove it.
Signed-off-by: Werner Koch <wk@gnupg.org>
* scd/scdaemon.h (opt): Add field opcsc_shared.
* scd/scdaemon.c (opcscShared): New.
(opts): Add "--pcsc-shared".
(main): Set flag.
* scd/apdu.c (connect_pcsc_card): Use it.
(pcsc_get_status): Take flag in account.
* scd/app-openpgp.c (cache_pin): Bypass in shared mode.
(verify_chv2: Do not auto verify chv1 in shared mode.
* scd/app-piv.c (cache_pin): By pass caceh in shared mode.
--
This option should in general not be used. The patch tries to limit
bad effects but using shared mode is somewhat dangerous depending on
the other PC/SC users.
* common/gc-opt-flags.h (GC_OPT_FLAG_RUNTIME): Move to ...
* tools/gpgconf-comp.c: here.
(known_options_scdaemon): Remove "options".
(known_options_dirmngr): Remove "options".
(known_options_gpgsm): Remove "options".
(known_options_gpg): Remove "options" and "keyserver".
(struct gc_option_s): Rename active t gpgconf_list.
(gc_component_list_options): Do not act upon active.
(option_check_validity): Ditto.
(is_known_option): Make it work correctly for unknown options.
(retrieve_options_from_program): Use renamed flag gpgconf_list only to
detect duplicated items from --gpgconf-list. Do not set runtime.
Only e set the options if set by --gpgconf-list; never clear them.
* agent/gpg-agent.c: Simplify the --gpgconf-list output.
* dirmngr/dirmngr.c: Ditto.
* g10/gpg.c: Ditto.
* kbx/keyboxd.c: Ditto.
* scd/scdaemon.c: Ditto.
* sm/gpgsm.c: Ditto.
* tests/openpgp/gpgconf.scm: Use "compliance" instead of "keyserver"
for the string arg test.
--
There is no need to read the list of options from the components
unless they convey a default value. It is better to consult only the
list we have in gpgconf-comp.c to decide on whether an option should
be displayed. Right, this might mess up thing if a newer gpgconf
version is used with an older component, but we already print warnings
in this case and in general we do not want to support this anymore -
the times of gpg 1.4. and 2.0 are long over now.
GnuPG-bug-id: 4788
Signed-off-by: Werner Koch <wk@gnupg.org>
* tools/gpgconf-comp.c: Remove all regular option descriptions. They
are now read in from the component. Also remove a few meanwhile
obsolete options.
* agent/gpg-agent.c: Add option description which were only set in
gpgconf-comp.c.
* dirmngr/dirmngr.c: Ditto.
* scd/scdaemon.c: Ditto.
* sm/gpgsm.c: Ditto.
* g10/gpg.c: Ditto.
--
This second part removes all regular option descriptions because they
can be read from the components. A few were missing in the components
and thus moved to there.
Signed-off-by: Werner Koch <wk@gnupg.org>
* scd/scdaemon.c: Switch to the new option parser and enable a global
conf file.
* dirmngr/dirmngr.c: Ditto.
* g13/g13.c: Ditto.
* g13/g13-syshelp.c: Ditto. Do not force verbose mode.
* dirmngr/dirmngr_ldap.c: Switch to the new option parser.
* dirmngr/dirmngr-client.c: Switch to the new option parser.
--
This finalizes the switch to the new option parser. What's left is to
remove the old argparser code from common.
g13-syshelp does not anymore default to --verbose because that can now
be enabled in /etc/gnupg/g13-syshelp.conf.
GnuPG-bug-id: 4788
Signed-off-by: Werner Koch <wk@gnupg.org>
* common/util.h: Remove argparse.h.
* common/argparse.c: Undef GPGRT_ENABLE_ARGPARSE_MACROS.
* configure.ac (GPGRT_ENABLE_ARGPARSE_MACROS): Define.
* agent/gpg-agent.c: Undef GPGRT_ENABLE_ARGPARSE_MACROS and include
argparse.h. Do this also for all main modules which use our option
parser except for gpg. Replace calls to strusage by calls to
gpgrt_strusage everywhere.
* g10/gpg.c (opts): Change type to gpgrt_opt_t. Flag oOptions and
oNoOptions with ARGPARSE_conffile and ARGPARSE_no_conffile.
(main): Change type of pargs to gpgrt_argparse_t. Rework the option
parser to make use of the new gpgrt_argparser.
--
This is not yet finished but a make check works. gpg has the most
complex and oldest option handling and thus this is the first
migration target. SE-Linux checks and version-ed config files are
missing and will be added later.
GnuPG-bug-id: 4788
Signed-off-by: Werner Koch <wk@gnupg.org>
No functional changes, just fixing minor spelling issues.
---
Most of these were identified from the command line by running:
codespell \
--ignore-words-list fpr,stati,keyserver,keyservers,asign,cas,iff,ifset \
--skip '*.po,ChangeLog*,help.*.txt,*.jpg,*.eps,*.pdf,*.png,*.gpg,*.asc' \
doc g13 g10 kbx agent artwork scd tests tools am common dirmngr sm \
NEWS README README.maint TODO
Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
* common/util.h (EXTERN_UNLESS_MAIN_MODULE): Add the definion only
here but now without the Norcroft-C. Change all other places where it
gets defined.
* common/iobuf.h (iobuf_debug_mode): Declare unconditionally as
extern.
* common/iobuf.c (iobuf_debug_mode): Define it here.
* agent/gpg-agent.c (INCLUDED_BY_MAIN_MODULE): Define here and also in
all main modules of all other programs.
* g10/main.h: Put util.h before the local header files.
--
This change is required for use with gcc/ld's LTO feature which does
not allow common blocks. Further gcc 10 will make -fno-common the
default and thus this chnage is always needed. What a pitty.
Co-authored-by: Tomáš Mráz
GnuPG-bug-id: 4831
Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit 21d9bd8b87)
- Applied respective chnages also to gpg-card and keyboxd.
Signed-off-by: Werner Koch <wk@gnupg.org>
* scd/scdaemon.h: Include app-common.h. Remove inclusion of that
header from all other files.
(card_t, app_t): Move typedef to ...
* scd/app-common.h: here. Use them in the defs.
--
In another patch we will need apptype_t in the ctrl object and thus we
need to reorganize things a bit now. Given that most files need
app-common anyway it makes sense to always include it.
Signed-off-by: Werner Koch <wk@gnupg.org>
* doc/scdaemon.texi (card-timeout): Remove.
* scd/scdaemon.c (main): Remove oCardTimeout handling.
--
There was the card-timeout option in GnuPG 2.0, but it was never
implemented correctly. The intention of this option was to allow
sharing smartcard among multiple applications, but this didn't work
well as user's expectation (it only worked with DISCONNECT command).
This is because other parts of scdaemon assumes exclusive access. In
GnuPG 2.1, the support of the option was removed, improving
"DISCONNECT" command always works well without this option.
GnuPG-bug-id: 3383
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/scdaemon.c (setup_signal_mask): New.
(main): Call setup_signal_mask.
(handle_connections): Remove signal mask setup.
--
For new thread, signal mask is inherited by thread creation.
Thus, it is best to setup signal mask just after npth_init.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* configure.ac (require_pipe_to_unblock_pselect): Default is "yes".
* scd/scdaemon.c (scd_kick_the_loop): Minor clean up.
--
Normally SIGCONT or SIGUSR2 works for unblocking pselect. But on my
machine with GNU/Linux, when a machine is suspend/resume-ed, pselect
keeps blocked, while signal itself is delivered.
It's better to use pipe.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/apdu.c (struct reader_table_s): Add field of current_state.
(new_reader_slot): Initialize current_state.
(pcsc_get_status): Keep the status in READER_TABLE array.
Return SW_HOST_NO_READER when PCSC_STATE_CHANGED.
* scd/scdaemon.c (handle_connections): Silence a warning.
--
To detect some change of card status, including suspend/resume
possibly, SCardGetStatusChange should be used keeping the
dwCurrentState field.
This change could improve situation for suspend/resume with Yubikey on
Windows. Even not, this is doing the Right Thing.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* configure.ac (HAVE_PSELECT_NO_EINTR): New.
* scd/scdaemon.c (scd_kick_the_loop): Write to pipe.
(handle_connections): Use pipe.
--
On NetBSD, signal to the same process cannot unblock pselect,
with unknown reason. Use pipe instead, for such systems.
GnuPG-bug-id: 3778
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/gpg-agent.c (oListenBacklog): New const.
(opts): New option --listen-backlog.
(listen_backlog): New var.
(main): Parse new options.
(create_server_socket): Use var instead of 5.
* dirmngr/dirmngr.c: Likewise.
* scd/scdaemon.c: Likewise.
--
GnuPG-bug-id: 3473
Signed-off-by: Werner Koch <wk@gnupg.org>
* scd/apdu.c (open_ccid_reader): Fix error handling of ccid_get_atr.
* scd/app.c (select_application): Always kick the loop if new APP.
* scd/ccid-driver.c (ccid_open_usb_reader): Don't setup at open.
(ccid_slot_status): Setup interrupt transfer when !ON_WIRE.
--
We can use the interrupt transfer to be notified about card status
change. In this case, we don't need to issue PC_to_RDR_GetSlotStatus
command. This change improve the setup the notification; it should be
done after registration of APP.
When the setup is done just after opening the USB connection (before
issuing PC_to_RDR_IccPowerOn), a reader might notifies about no card
availability (because of not yet powered on), even though the card is
ready to be powered on.
GnuPG-bug-id: 3508
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/gpg-agent.c (main): Always to the chdir.
* dirmngr/dirmngr.c (main): Ditto.
* scd/scdaemon.c (main): Ditto.
--
Note that only dirmngr did not call the chdir with --no-detach. thus
we kept it this way.
Tested gpg-agent by checking the properties shown by procexp.
Gnupg-bug-id: 2670
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/gpg-agent.c [HAVE_W32_SYSTEM]: Include <direct.h>.
(main) [HAVE_W32_SYSTEM]: Change working directory to \.
* dirmngr/dirmngr.c [HAVE_W32_SYSTEM]: Include <direct.h>.
(main) [HAVE_W32_SYSTEM]: Change working directory to \.
* scd/scdaemon.c [HAVE_W32_SYSTEM]: Include <direct.h>.
(main) [HAVE_W32_SYSTEM]: Change working directory to \.
Signed-off-by: Marcus Brinkmann <mb@g10code.com>
GnuPG-bug-id: 2670
* scd/scdaemon.c (notify_fd): Remove.
(the_event) [W32]: New.
(main_thread_pid) [!W32]: New.
(handle_signal): Handle SIGCONT.
(scd_kick_the_loop): Use signal on UNIX and event on Windows.
(handle_connections): Likewise.
--
Code with CreateEvent is copied from gpg-agent.c.
Code for signal is copied from dkg's gpg-agent-idling in Debian.
GnuPG-bug-id: 2982
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/scdaemon.c (get_active_connection_count): New.
(start_connection_thread): Bump ACTIVE_CONNECTIONS up and down.
* scd/command.c (cmd_getinfo): Add subcommand "connections".
--
Apply gpg-agent change to scdaemon. See the commit in 2016-08-06:
40d16029ed
Then, add kicking the loop, so that main loop can notice the change of
the connection.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/scdaemon.c (ticker_disabled): Remove.
(handle_tick, need_tick): Remove.
(handle_connections): Don't check ticker_disabled.
--
Now, removal of device/card is only done by the function
scd_update_reader_status_file, it should be called if needed.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/app.c (select_application): Fix the logic if periodical check is
needed. If it is needed for newly found device(s), kick the loop.
(scd_update_reader_status_file): Return value if select(2) should be
called with timeout.
* scd/ccid-driver.c (ccid_require_get_status): Don't return 0 for
token with no interrupt transfer for now.
* scd/command.c (open_card_with_request): Fix scan by SERIALNO.
* scd/scdaemon.c (update_usb): Remove.
(handle_connections): Evaluate need_tick after handle_tick.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/app-common.h (struct app_ctx_s): Rename field to
periodical_check_needed.
* scd/scdaemon.c (update_usb): Rename from update_fdset_for_usb.
Don't use libusb_get_pollfds any more.
(scd_kick_the_loop): New.
(need_tick): Follow the rename.
(handle_connections): No libusb event handling here.
* scd/app.c (app_new_register): Follow the change of rename.
(select_application, scd_update_reader_status_file): Likewise.
* scd/ccid-driver.c (ccid_usb_thread_is_alive): New.
(intr_cb): Call scd_kick_the_loop.
(ccid_usb_thread): New. Thread to invoke INTERRUPT callback.
(ccid_open_usb_reader): Add thread invocation.
(ccid_require_get_status): Remove
LIBUSB_WORKS_EXPECTED_FOR_INTERRUPT_ENDP.
(do_close_reader): Carefully handle handle->transfer.
(get_escaped_usb_string): Insert npth_unprotect/npth_protect.
(do_close_reader, bulk_out, bulk_in, abort_cmd, ccid_slot_status)
(ccid_transceive, ccid_transceive_secure): Likewise.
--
It found that libusb_get_pollfds is not supported on Windows.
Besides, it's a bit difficult to use for the select loop.
Thus, we use the thread named ccid_usb_thread, instead.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/apdu.c(struct reader_table_s): Add require_get_status.
(apdu_connect): Change return value meaning. Call apdu_reset here.
* scd/app.c (app_new_register): Add require_get_status.
(select_application): Use the return value of apdu_connect.
(scd_update_reader_status_file): Call update_fdset_for_usb with
checking all_have_intr_endp.
(app_list_start, app_list_finish): Remove.
* scd/ccid-driver.c (struct ccid_driver_s): Add transfer.
(intr_cb): Don't call libusb_transfer in this callback.
(ccid_require_get_status): New.
(do_close_reader): Call libusb_transfer here.
* scd/scdaemon.c (update_fdset_for_usb): Remove the first argument.
--
With Gnuk Token, it works fine as expected. With Gemalto reader,
intr_cb is not called when card is removed. So, the macro
LIBUSB_WORKS_EXPECTED_FOR_INTERRUPT_ENDP is not defined yet.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/scdaemon.c (update_fdset_for_usb): Wake up the select(2).
(handle_connections): Use a kind of "self-pipe" technique.
--
Use pipe to wake up select(2). If UNIX-only, signal could be used. For
portability, "self-pipe" is better, here. Setup for non-blocking for
pipe fds are not needed, because speed of USB device insertion is
limited by human physical interaction; No one can do hundreds of
device insertion/removal-s per second.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/apdu.c (reader_table_lock, apdu_init): New.
(new_reader_slot): Serialize by reader_table_lock.
* scd/app.c (lock_app, unlock_app, app_new_register): Fix error code
usage.
(initialize_module_command): Call apdu_init.
* scd/scdaemon.c (main): Handle error for initialize_module_command.
--
This is a long standing bug. There are two different things; The
serialization of allocating a new SLOT, and the serialization of using
the SLOT. The latter was implemented in new_reader_slot by lock_slot.
However, the former was not done. Thus, there was a possible race where
a same SLOT is allocated to multiple threads.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* dirmngr/dirmngr.c (main): Before calling setsid do not close the
standard fds but connect them to /dev/null.
* scd/scdaemon.c (main): Ditto. Noet that the old test for a log
stream was even reverted.
--
Note that this was fixed for gpg-agent 10 years ago on 2006-11-09.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/gpg-agent.c (thread_init_once, initialize_modules): New.
(main): Make sure no daemonizing-fork call after npth_init, and no npth
calls before npth_init, with care of npth calls by assuan hooks.
* dirmngr/dirmngr.c (thread_init): New.
(main): Make sure npth_init must not be called before daemonizing fork.
* scd/scdaemon.c (main): Likewise.
--
It is simply the best for nPth not to allow the daemonizing fork after
npth_init, because semantics and implementations of forked child process
in a threaded application is a difficult corner case.
GnuPG-bug-id: 1779
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
