* g10/tdbio.c (tdbio_read_record): Return GPG_ERR_EOF.
(tdbio_new_recnum): Never return on error.
(lookup_hashtable): Print a more descriptive error in case of !TABLE.
--
Also: tdbio_new_recnum had a bug in that it returned an error code and
not a record number in the error case. The function is expected to
always return a valid new record number.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/tdbio.c (is_locked): Turn into a counter.
(take_write_lock, release_write_lock): Implement recursive locks.
--
On trustdb creation we have this call sequence:
init_trustdb -> takes lock
tdbio_set_dbname
create_version_record
tdbio_write_record
put_record_into_cache -> takes lock
put_record_into_cache -> releases lock
init_trustdb -> releases lock
The second take lock does noting but the first release lock has
already released the lock and the second release lock is a thus a NOP.
This is likely the cause for the corrupted trustdb as reported in
GnuPG-bug-id: 3839
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/tdbio.c (in_transaction): Comment this var.
(put_record_into_cache): Comment the transaction code.
(tdbio_sync): Ditto
Signed-off-by: Werner Koch <wk@gnupg.org>
* sm/server.c: Include shareddefs.h.
(option_handler): Add option.
--
This is required when running gpgsm in server mode as done by GPGME.
Noet that a command line option takes precedence.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/gpg.c (oRequestOrigin): New const.
(opts): New option --request-origin.
(main): Parse that option.
* g10/options.h (struct opt): Add field request_origin.
* g10/call-agent.c (start_agent): Send option to the agent.
* sm/gpgsm.c (oRequestOrigin): New const.
(opts): New option --request-origin.
(main): Parse that option.
* sm/gpgsm.h (struct opt): Add field request_origin.
* sm/call-agent.c (start_agent): Send option to the agent.
Signed-off-by: Werner Koch <wk@gnupg.org>
* common/shareddefs.h (request_origin_t): New.
* common/agent-opt.c (parse_request_origin): New.
(str_request_origin): New.
* agent/command.c (option_handler): Implement new option.
--
This allows to pretend that a request originated from the extra or
browser socket.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/parse-packet.c (enum_sig_subpkt): Check buflen before reading
the type octet. Print diagnostic.
--
If the final subpacket has only a length header evaluating to zero and
missing the type octet, a read could happen right behind the buffer.
Valgrind detected this. Fix is obvious. Note that the further
parsing of the subpacket is still okay because it always checks the
length. Note further that --list-packets uses a different code path
and already reported an error.
Reported-by: Philippe Antoine
He provided a test file copied below. Running "gpg -v --verify" on it
triggered the bug.
-----BEGIN PGP ARMORED FILE-----
Comment: Use "gpg --dearmor" for unpacking
kA0DAAoBov87N383R0QBrQJhYgZsb2wucHlaqTVWYnl0ZXMgPSBbMHg1LCAweDY0
LCAweDRjLCAweGM0LCAweDMsIDB4MCwgMHg0LCAweDAsIDB4YWMsIDB4YSwgMHhj
MSwgMHhjMSwgMHgyLCAweDEsIDB4MiwgMHg3LCAweDQwLCAweDIsIDB4MiwgMHgy
LCAweDIsIDB4MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAweDJkLCAweGRkLCAweDIs
IDB4MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAweDIsIDB4
MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAweDIsIDB4NzcsIDB4ODcsIDB4MiwgMHgy
LCAweDIsIDB4MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAw
eDIsIDB4MiwgMHgyLCAweDIsIDB4MiwgMHg3NywgMHg4NywgMHgyLCAweDIsIDB4
MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAweDIsIDB4MiwgMHgyLCAweDIsIDB4Miwg
MHgyLCAweDIsIDB4MiwgMHgyLCAweDc3LCAweDg3LCAweDIsIDB4MiwgMHgyLCAw
eDIsIDB4MiwgMHgyLCAweDIsIDB4MCwgMHhhZF0KCmZvciBpIGluIHJhbmdlKGxl
bihieXRlcykpOgogICAgaWYgaSUxNiA9PSAwOgogICAgICAgIHByaW50CiAgICAg
ICAgcHJpbnQgIiUwNngiICUgaSwKICAgIHByaW50ICIlMDJ4ICIgJSBieXRlc1tp
XSwKiQJNBAABCgAeFiEEU+Y3aLjDLA3x+9Epov87N383R0QFAlqpNVYAAAoJEKL/
Ozd/N0dElccP/jBAcFHyeMl7kop71Q7/5NPu3DNULmdUzOZPle8PVjNURT4PSELF
qpJ8bd9PAsO4ZkUGwssY4Kfb1iG5cR/a8ADknNd0Cj9/QA2KMVNmgYtReuttAjvn
hQRm2VY0tvDCVAPI/z8OnV/NpOcbk8kSwE+shLsP7EwqL5MJNMXKqzm1uRxGNYxr
8TNuECo3DO64O2NZLkMDXqq6lg+lSxvDtXKxzKXgVC+GMtOE56lDwxWLqr39f9Ae
Pn0q2fVBKhJfpUODeEbYSYUp2hhmMUIJL/ths9MvyRZ9Z/bHCseFPT58Pgx6g+MP
q+iHnVZEIVb38XG+rTYW9hvctkRZP/azhpa7eO8JAZuFNeBGr4IGapwzFPvQSF4B
wBXBu0+PPrV9VJVe98P4nw2xcuJmkn6mgZhRVYSqDIhY64bSTgQxb/pdtGwrTjtL
WoUKVI+joLRPnDmwexH9+QJCB+uA6RsN/LqsQfDseyr40Z6dHJRqWGgP3ll6iZgw
WF768uiIDJD8d4fegVnkpcH98Hm0I/dKsMR1MGV/sBxYC8mAOcOWwSPNGDqPlwwR
eWPdr1O6CoYEWwiZMicSe0b5TsjB5nkAWMy7c9RyhtMJzCQ/hFpycpj0A0Zs+OGa
eJQMZZV0s8AQZ04JzoX0zRpe0RcTyJn3Tr6QGbVi9tr+QdKHFuDMUqoX
=qYZP
-----END PGP ARMORED FILE-----
Signed-off-by: Werner Koch <wk@gnupg.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>
* g10/card-util.c (factory_reset): Dummy PIN size is now 32-byte.
Connect the card again at the last step.
--
Before the change, a user has to quit the session to continue. Now,
it is possible to type RET in the session and see if it's really done.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* scd/ccid-driver.c (intr_cb): Try submitting INTERRUPT urb
to see if it's suspend/resume.
--
Upon suspend/resume, LIBUSB_TRANSFER_NO_DEVICE is returned, since all
URBs are cancelled. We need to see if it's real NODEV error or its by
suspend/resume. We can distinguish by sending URB again.
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>
--
WIN32_LEAN_AND_MEAN is required to avoid definitions of grp1, grp2,
and grp3 in dlgs.h, which is included by windows.h.
Fixes-commit: fd595c9d3642dba437fbe0f6e25d7aaaae095f94
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/cache.c (agent_cache_housekeeping): New func.
* agent/gpg-agent.c (handle_tick): Call it.
--
This change mitigates the risk of having cached items in a post mortem
dump.
GnuPG-bug-id: 3829
Signed-off-by: Werner Koch <wk@gnupg.org>
--
Sorry, I accidentally pushed the last commit without having amended it
with this fix.
Fixes-commit: fd595c9d3642dba437fbe0f6e25d7aaaae095f94
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/call-agent.h (agent_card_info_s): Add fields grp1, grp2 and
grp3.
* g10/call-agent.c (unhexify_fpr): Allow for space as delimiter.
(learn_status_cb): Parse KEYPARIINFO int the grpX fields.
* g10/card-util.c (print_keygrip): New.
(current_card_status): Print "grp:" records or with --with-keygrip a
human readable keygrip.
--
Suggested-by: Peter Lebbing <peter@digitalbrains.com>
Signed-off-by: Werner Koch <wk@gnupg.org>
* tools/gpgconf-comp.c (get_config_filename): Allow UNC paths.
--
The homedir of GnuPG on Windows can be on a network share
e.g. if %APPDATA% is redirected to a network share. The
file API calls work and GnuPG itself works nicely
with such paths so gpgconf should work with them, too.
GnuPG-Bug-Id: T3818
Signed-off-by: Andre Heinecke <aheinecke@intevation.de>
* dirmngr/ks-engine-hkp.c (handle_send_request_error): Handle two more
error codes.
--
Handle the two possible connect failures related to missing IPv6 support
gracefully by marking the host dead and retrying with another one.
If IPv6 is disabled via procfs, connect() will return EADDRNOTAVAIL.
If IPv6 is not compiled into the kernel, it will return EAFNOSUPPORT.
This makes it possible to use dual-stack hkp servers on hosts not having
IPv6 without random connection failures.
GnuPG-bug-id: 3331
--
The above description seems to be for Linux, so it is possible that
other systems might behave different. However, it is worth to try
this patch.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/keydb.c (parse_keyblock_image): Use log_info instead of
log_error for skipped packets.
* g10/keyring.c (keyring_get_keyblock): Ditto.
--
log_info should be sufficient and makes this more robust. Some
tools (e.g. Enigmail) are too picky on return codes from gpg.
Signed-off-by: Werner Koch <wk@gnupg.org>
* tools/gpg-wks-client.c (get_key_status_parm_s)
(get_key_status_cb, get_key): Move to ...
* tools/wks-util.c: ...here.
(get_key): Rename to wks_get_key.
* tools/gpg-wks-server.c: Include userids.h.
(command_install_key): Allow use of a fingerprint.
Signed-off-by: Werner Koch <wk@gnupg.org>
* tools/gpg-wks.h (policy_flags_s): Add field 'submission_address'.
* tools/wks-util.c (wks_parse_policy): Parse that field.
(wks_free_policy): New.
* tools/gpg-wks-client.c (command_send): Also try to take the
submission-address from the policy file. Free POLICY.
* tools/gpg-wks-server.c (process_new_key): Free POLICYBUF.
(command_list_domains): Free POLICY.
Signed-off-by: Werner Koch <wk@gnupg.org>
--
This can be used to build GnuPG with static versions of the core
gnupg libraries. For example:
make -f build-aux/speedo.mk STATIC=1 SELFCHECK=0 \
INSTALL_PREFIX=/somewhere/gnupg22 native
The SELFCHECK=0 is only needed to build from a non-released version.
You don't need it with a released tarball.
Signed-off-by: Werner Koch <wk@gnupg.org>
* kbx/keybox-search.c (blob_cmp_fpr): Avoid overflow in OFF+LEN
checking.
(blob_cmp_fpr_part): Ditto.
(blob_cmp_name): Ditto.
(blob_cmp_mail): Ditto.
(blob_x509_has_grip): Ditto.
(keybox_get_keyblock): Check OFF and LEN using a 64 bit var.
(keybox_get_cert): Ditto.
--
On most 32 bit systems size_t is 32 bit and thus the check
size_t cert_off = get32 (buffer+8);
size_t cert_len = get32 (buffer+12);
if (cert_off+cert_len > length)
return gpg_error (GPG_ERR_TOO_SHORT);
does not work as intended for all supplied values. The simplest
solution here is to cast them to 64 bit.
In general it will be better to avoid size_t at all and work with
uint64_t. We did not do this in the past because uint64_t was not
universally available.
GnuPG-bug-id: 3770
Signed-off-by: Werner Koch <wk@gnupg.org>
* sm/export.c (sexp_to_kparms): Fix the computation of array[6],
which must be 'd mod (q-1)' but was 'p mod (q-1)'.
--
This bug is not serious but makes some consistency checks fail.
For example, 'openssl rsa -check' reports the following error:
$ gpgsm --out my.key --export-secret-key-raw 0xXXXXXXXX
$ openssl rsa -check -noout -inform DER -in my.key
RSA key error: dmq1 not congruent to d
--
Let me(wk) add this:
This bug was introduced with
Fixes-commit: 91056b1976bfb7b755e53b1302f4ede2b5cbc05d
right at the start of GnuPG 2.1 in July 2010. Before that (in 2.0) we
used gpg-protect-tool which got it right. We probably never noticed
this because gpgsm, and maybe other tools too, fix things up during
import.
Signed-off-by: Werner Koch <wk@gnupg.org>
* common/asshelp2.c (vprint_assuan_status_strings): New.
(print_assuan_status_strings): New.
* agent/command.c (agent_write_status): Replace by call to new
function.
* dirmngr/server.c (dirmngr_status): Ditto.
* g13/server.c (g13_status): Ditto.
* g13/sh-cmd.c (g13_status): Ditto.
* sm/server.c (gpgsm_status2): Ditto.
* scd/command.c (send_status_info): Bump up N.
--
This fixes a potential overflow if LFs are passed to the status
string functions. This is actually not the case and would be wrong
because neither the truncating in libassuan or our escaping is not the
Right Thing. In any case the functions need to be more robust and
comply to the promised interface. Thus the code has been factored out
to a helper function and N has been bumped up correctly and checked in
all cases.
For some uses this changes the behaviour in the error case (i.e. CR or
LF passed): It will now always be C-escaped and not passed to
libassuan which would truncate the line at the first LF.
Reported-by: private_pers
* scd/app-openpgp.c (get_cached_data): Return NULL for Data Object
with no data.
--
When GET_DATA returns no data with success (90 00), this routine
firstly returned buffer with length zero, and secondly (with cache)
returned NULL, which is inconsistent. Now, it returns NULL for both
cases.
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>
