* dirmngr/workqueue.c: New.
* dirmngr/Makefile.am (dirmngr_SOURCES): Add new file.
* dirmngr/server.c (server_local_s): New field session_id.
(cmd_wkd_get): Add a task.
(task_check_wkd_support): New stub function.
(cmd_getinfo): New sub-commands "session_id" and "workqueue".
(start_command_handler): Add arg session_id and store it in
SERVER_LOCAL.
(dirmngr_status_helpf): New.
* dirmngr/dirmngr.h (wqtask_t): New type.
* dirmngr/dirmngr.c (main): Pass 0 as session_id to
start_command_handler.
(start_connection_thread): Introduce a session_id and pass it to
start_command_handler. Run post session tasks.
(housekeeping_thread): Run global workqueue tasks.
--
Signed-off-by: Werner Koch <wk@gnupg.org>
* dirmngr/domaininfo.c (MAX_DOMAINBUCKET_LEN): New.
(insert_or_update): Limit the length of a bucket chain.
(domaininfo_print_stats): Print just one summary line.
Signed-off-by: Werner Koch <wk@gnupg.org>
* dirmngr/domaininfo.c: New file.
* dirmngr/Makefile.am (dirmngr_SOURCES): Add file.
* dirmngr/server.c (cmd_wkd_get): Check whether the domain is already
known and tell domaininfo about the results.
--
This adds a registry for domain information to eventually avoid
useless queries for domains which do not support WKD. The missing
part is a background task to check whether a queried domain supports
WKD at all and to expire old entries.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/gpg-agent.c (handle_connections): Always check inotify fds.
--
I noticed a gpg-agent processed, probably in shutdown_pending state,
which was selecting on only these two inotify fds. The select
returned immediately but because we did not handle the fds in
shutdown_pending state they were not read and the next select call
returned one of them immediately again. Actually that should not
hanppen because the
if (active_connections == 0)
break; /* ready */
should have terminated the loop. For unknown reasons (maybe be just a
connection thread terminated in a gdb session) that did not happen.
By moving the check outside of the shutdown_pending condition and
closing the fd after they have been triggered the code should be more
robust.
Signed-off-by: Werner Koch <wk@gnupg.org>
* tests/openpgp/Makefile.am [DISABLE_REGEX] (EXTRA_DIST, XTESTS):
Conditionalize.
* tests/openpgp/all-tests.scm (all-tests): Input file is Makefile.
--
The feature is only valid with !DISABLE_REGEX.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* tests/openpgp/Makefile.am (XTESTS): Add trust-pgp-4.scm.
(EXTRA_DIST): Remove the test file from EXTRA_DIST.
--
Now that issue 2923 is fixed, the trust-pgp-4 test passes as
expected and we can enable it again. That should help prevent
a future regression on this issue.
Signed-off-by: Damien Goutte-Gattat <dgouttegattat@incenp.org>
--
In https://bugs.debian.org/881393 , Jonas Smedegaard reports:
> In option number 1, the word "komprimeret" means "compressed".
>
> I am pretty sure it should say "kompromitteret" instead, which means
> "compromised".
Debian-Bug-Id: 881393
Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
* g10/trustdb.c (sanitize_regexp): Only escape operators.
--
To sanitize a regular expression, quoting by backslash should be only
done for defined characters. POSIX defines 12 characters including
dot and backslash.
Quoting other characters is wrong, in two ways; It may build an
operator like: \b, \s, \w when using GNU library. Case ignored match
doesn't work, because quoting lower letter means literally and no
much to upper letter.
GnuPG-bug-id: 2923
Co-authored-by: Damien Goutte-Gattat <dgouttegattat@incenp.org>
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/protect.c (calibrate_get_time): Use clock or clock_gettime.
--
For calibration, clock(3) is better than times(3) among UNIXen.
Tested on NetBSD 7.1 and FreeBSD 11.1, using QEMU.
Thanks to Damien Goutte-Gattat for the information of use of
CLOCKS_PER_SEC; The old code with times(3) is not 100% correct,
in terms of POSIX. It should have used sysconf (_SC_CLK_TCK) instead
of CLOCKS_PER_SEC. CLOCKS_PER_SEC is specifically for clock(3).
GnuPG-bug-id: 3056, 3276, 3472
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* tests/openpgp/defs.scm (create-gpghome): Add s2k-count.
--
My tests show only 2.5% improvement, but as we have that option now
let's use it.
real 9m12.604s
user 2m20.720s
sys 0m11.452s
real 8m3.815s
user 2m16.700s
sys 0m11.544s
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/command.c (cmd_getinfo): New sub-commands.
* agent/protect.c (get_standard_s2k_count): Factor some code out to ...
(get_calibrated_s2k_count): new.
(get_standard_s2k_time): New.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/agent.h (opt): New field 's2k_count'.
* agent/gpg-agent.c (oS2KCount): New enum value.
(opts): New option --s2k-count.
(parse_rereadable_options): Set opt.s2k_count.
--
This option is useful to speed up the starting of gpg-agent and in
cases where the auto-calibration runs into problems due to a broken
time measurement facility.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/card-util.c (ask_card_keyattr): Print "rsaNNNN".
--
This is a separate patch from the previous to avoid string changes
when backporting the other patch to 2.2.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/card-util.c (show_keysize_warning): Slightly change the text.
(ask_card_keyattr): Handle special value 25519.
(do_change_keyattr): Allow changing to cv25519/ed25519.
(generate_card_keys): Ditto.
(card_generate_subkey): Ditto.
--
This is kludge to make it easier for gnuk to be switched into ECC
mode.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/card-util.c (ask_card_rsa_keysize): Rename to ask_card_keyattr.
(do_change_rsa_keysize): Rename to do_change_keyattr.
--
We want to support other algos than RSA and thus we need a better name
for the functions.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/card-util.c (factory_reset): Simplify.
--
In this summer, I got report about old code before this change didn't
work with newer Yubikey. I got another report test version of OpenPGP
card V3.3 implementation didn't work, either. Then, I confirmed that
according to the OpenPGP card specification, the procedure of old code
is not expected by its author.
This change simplify "factory-reset" as simple.
Only versions of Gnuk 1.2.2, 1.2.3, 1.2.4, won't work with this
change. That's because the factory-reset feature of Gnuk was
introduced by reading the implementation of GnuPG, instead of reading
the specification. Gnuk 1.2.5 and later works well. All OpenPGPcard
implementations I have work well (2.0, 2.1, 2.2, test version of 3).
GnuPG-bug-id: 3286
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/agent.h (struct server_control_s): Rename PINENTRY_ACTIVE.
* agent/call-pinentry.c (entry_owner): Remove.
(agent_reset_query): Use thread private object of PINENTRY_ACTIVE.
(unlock_pinentry): Add CTRL to arguments to access thread private.
Check and decrement PINENTRY_ACTIVE for recursive use.
(start_pinentry): Check and increment PINENTRY_ACTIVE for recursion.
(agent_askpin): Follow the change of unlock_pinentry API.
(agent_get_passphrase, agent_get_confirmation): Likewise.
(agent_show_message, agent_popup_message_start): Likewise.
(agent_popup_message_stop, agent_clear_passphrase): Likewise.
--
We use the member PINENTRY_ACTIVE as a thread private object.
It's only valid for a single thread at a time.
It would be possible to have a thread shared object of
PINENTRY_ACTIVE, keeping ENTRY_OWNER for distinguishing its
owner (which is also a thread shared object). But, in this case,
access to ENTRY_OWNER is tricky (only comparison to accessing thread
would be OK with no lock), or we need to introduce another lock for
accessing ENTRY_OWNER, which complicates the code too much.
So, simply have a thread private object for recursive pinentry access.
GnuPG-bug-id: 3190
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/command.c (cmd_scd): Support !BUILD_WITH_SCDAEMON.
* tests/openpgp/defs.scm (create-gpghome): Likewise.
* tests/gpgsm/gpgsm-defs.scm (create-gpgsmhome): Likewise.
--
We could modify gpg-agent to remove all support of scdaemon, with no
inclusion of call-scd.c, divert-scd.c, and learncard.c, but it would
not be worth to do that.
GnuPG-bug-id: 3316
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/command.c (cmd_scd): Support !BUILD_WITH_SCDAEMON.
* tests/openpgp/defs.scm (create-gpghome): Likewise.
* tests/gpgsm/gpgsm-defs.scm (create-gpgsmhome): Likewise.
--
We could modify gpg-agent to remove all support of scdaemon, with no
inclusion of call-scd.c, divert-scd.c, and learncard.c, but it would
not be worth to do that.
GnuPG-bug-id: 3316
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* g10/key-check.c (print_info): New.
(key_check_all_keysigs): Print sig checking results only in debug
mode. Prettify the stats info and suppress them in quiet mode.
--
This also makes usable stats by prefixing them with the key and the
program name.
GnuPG-bug-id: 3397
Signed-off-by: Werner Koch <wk@gnupg.org>
* configure.ac: Implement that option.
--
This can be used as a workaround in case of bogus autoconf tests.
GnuPG-bug-id: 2423
Signed-off-by: Werner Koch <wk@gnupg.org>
* configure.ac: Do not mess with the user provided CFLAGS.
--
A problem was claimed with some configure tests if the user provided
CFLAGS=-Werror. The commit introducing this
Fixes-commit: 02eb9fc9d5863abcfed6af704e618f8cac7cc2e8
does not mention a concrete case. Anyway, messing with CFLAGS is a
bad idea because configure tests will then test something different
than what is used later (cf. autoconf manual). Tests which depend on
the whether -Werror is used needsto be fixed.
Note that in certain cases we modify CFLAGS. This is only done for
some configure options or if the platform requires the use of special
compiler flags (e.g. on HP/UX).
GnuPG-bug-id: 2423
* sm/certlist.c (gpgsm_find_cert): Add arg allow_ambiguous and use it.
* sm/call-dirmngr.c (inq_certificate): Pass true to ALLOW_AMBIGUOUS
(run_command_inq_cb): Ditto.
* sm/gpgsm.c (main): Pass false.
* sm/server.c (cmd_passwd): Pass false.
--
As described in my report T1644, it is possible that multiple
certificates exist with the same Distinguished Name and the same key.
In this case, verifying S/MIME signatures and other actions fail with
"certificate not found: Ambiguous name". For details see the bug
report.
To circumvent the problem, I am patching GnuPG since 2014 so that in
this case the newest of the ambiguous certificates is used.
This is not an ultimate solution of the problem: You should try every
certificate with the same DN until verification succeeds or until all
certificates fail, and if multiple certificates of a chain are
ambiguous you even have to check every combination. You may even
consider checking the keyUsage attributes of the ambiguous certificates
to reduce the number of combinations.
But in the existing case of the certificates in the German Research
Network (DFN) PKI where the newest one is the valid one and all
ambiguous certificates have the same keyUsage attributes, this patch
has proven to be sufficient over the last three years.
With every GnuPG update, I have adapted the patch, luckily I never
needed to change anything except line numbers.
GnuPG-bug-id: 1644
ChangeLog log written by wk, comment taken from mail. Signed-off line
was missing in the plain diff. However the mail with the patch and
the DCO posted as reply to that mail were both signed.
Signed-off-by: Werner Koch <wk@gnupg.org>
* tools/gpgconf-comp.c (retrieve_options_from_program): Add arg
only_installed.
(gc_component_retrieve_options): Use this if we want to process all
components.
--
Note that this also also ignores them in --with-defaults. This is
useful for systems which come without scdaemon.
GnuPG-bug-id: 3313
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/keyedit.c (KEYEDIT_NEED_SUBSK): New.
(cmds): Add this flag to keytocard, bkuptocard, expire, and passwd.
(keyedit_menu): Check whether only subkeys are available and take care
of that in the command check and in the HELP listing. Also print a
different notice if only subkeys are available.
--
Print "Secret key is available" and the bailing out in all commands
which require the _primary_ secret key was surprising. Now we print
another notice and adjust the checks.
GnuPG-bug-id: 3463
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/keyedit.c (KEYEDIT_NOT_SK, KEYEDIT_ONLY_SK): Remove.
(cmds): Remove them.
--
These flags were cruft from the time we had to switch between secret
and public key view.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/keygen.c (generate_subkeypair): Ignore error code issued for
trying to verify a card based key.
--
We try to verify the primary key and thus seed the passphrase cache
before generating the subkey. However, the verification does not yet
work for on-card keys and thus the PASSWD --verify send to the agent
returns an error. This patch detects this error and continues without
a seeded passphrase cache. After all that pre-seeding is just a
convenience.
GnuPG-bug-id: 3280
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/import.c (import_one): Pass FROM_SK to list_keyblock_direct.
--
Note that this will likely add the suffix '#' top "sec" because the
secret key has not yet (or will not be) imported. If the secret key
already exists locally another suffix might be printed. The upshot is
that the suffix has no usefulness.
GnuPG-bug-id: 3431
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/import.c (import_secret_one): Check for dry-run before
transferring keys.
--
The use of --dry-run or --import-option show-only had no effect when
importing a secret key and the public key already existed. If the
public key did not exist an error message inhibited the import of the
secret key.
Signed-off-by: Werner Koch <wk@gnupg.org>
* dirmngr/ks-engine-http.c (ks_http_fetch): Forbid redirects from
a https URI to a http URI.
--
GnuPG-bug-id: 3436
Signed-off-by: Damien Goutte-Gattat <dgouttegattat@incenp.org>
* g10/pkclist.c (find_and_check_key): Call get_validity on a specific
keyblock.
--
When we have multiple keyrings, get_validity after
get_best_pubkey_byname should access same keyring. Or else, the
situation of an expired key in keyring A but valid key in keyring B
causes SEGV.
Thanks to Guido Günther for the use case and the log.
Debian-bug-id: 878812
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* g10/keydb.c (keydb_handle): New field 'keep_lock'.
(keydb_release): Clear that flag.
(keydb_lock): New function.
(unlock_all): Skip if KEEP_LOCK is set.
* g10/getkey.c (get_keyblock_byfprint_fast): Call keep_lock if
requested.
--
That change is straightforward. It helps to avoid the race condition
that another gpg process inserts a key while the first process is
between the search and the insert.
A similar change is due for gpgsm.
Note that the key edit operations may still suffer from a race.
GnuPG-bug-id: 3446
* g10/getkey.c (get_pubkey_byfprint_fast): Factor most code out to ...
(get_keyblock_byfprint_fast): .. new function.
* g10/import.c (revocation_present): s/int rc/gpg_error_t err/.
(import_one): Use get_keyblock_byfprint_fast to get the keyblock and a
handle. Remove the now surplus keyblock fetch in the merge branch.
Signed-off-by: Werner Koch <wk@gnupg.org>
* g10/import.c (import_keys_internal): Return gpg_error_t instead of
int. Change var names.
(import_keys_es_stream): Ditto.
(import_one): Ditto. Use a single keydb_new and simplify the use of
of keydb_release.
--
Note that this opens a keydb handle before we call
get_pubkey_byfprint_fast which internally uses another key db handle.
A further patch will cleanup this double use. Note that we also
disable the keydb caching for the insert case.
The s/int/gpg_error_t/ has been done while checking the call chains of
the import functions and making sure that gpg_err_code is always used.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/agent.h (server_control_s): Add field 'client_uid'.
* agent/call-pinentry.c (start_pinentry): Add uid field to assuan
option "owner" sent to pinentry.
* agent/command-ssh.c (peer_info_s): New static struct.
(get_client_pid): Rename to...
(get_client_info): Here, and extract uid in addition to pid.
(start_command_handler_ssh): Use get_client_info() instead of
get_client_pid().
* agent/command.c (start_command_handler): Try assuan_get_peercred,
and only fall back to assuan_get_pid when assuan_get_peercred fails.
--
This also requires an update to pinentry to handle the new uid field.
Distributing the uid as well as the pid makes it harder for a
different user on the same machine to take advantage of any race
conditions between when a requesting process might ask for something
that needs pinentry, and when pinentry gets around to inspecting the
state of that process.
We put the uid before the nodename because the uid is guaranteed to be
a integer (represented in decimal), which makes it much simpler to
parse past than the potentially arbitrarily structured nodename.
Use a / instead of whitespace to delimit pid/uid at Werner's request.
If we were willing to depend on the nodename being
whitespace-delimited (as the current, unreleased pinentry code does),
then we could add the uid after the nodename. But since no released
pinentry depends on this option anyway, i think we should make the
more conservative, easily-parseable choice and put the user ID first.
Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
