gnupg2.git
3 years agoagent: Support --mode=ssh option for CLEAR_PASSPHRASE.
NIIBE Yutaka [Fri, 25 Jan 2019 03:08:09 +0000 (12:08 +0900)]
agent: Support --mode=ssh option for CLEAR_PASSPHRASE.

* agent/command.c (cmd_clear_passphrase): Add support for SSH.

--

GnuPG-bug-id: 4340
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
(cherry picked from commit ae966bbe9b16ed68a51391afdde615339755e22d)
(cherry picked from commit 77a285a0a94994ee9b42289897f9bf3075c7192d)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name agent-Support-mode-ssh-option-for-CLEAR_PASSPHRASE.patch

3 years agogpgv: Improve documentation for keyring choices
Daniel Kahn Gillmor [Sun, 3 Mar 2019 15:22:34 +0000 (10:22 -0500)]
gpgv: Improve documentation for keyring choices

* doc/gpgv.texi: Improve documentation for keyring choices

--

From the existing documentation, it's not clear whether the default
keyring will always be mixed into the set of keyrings, or whether it
will be skipped if a --keyring is present.  The updated text here
attempts to describe the keyring selection logic more completely.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
(cherry picked from commit a7b2a87f940dba078867c44f1f50d46211d51719)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name gpgv-Improve-documentation-for-keyring-choices.patch

3 years agosm: Don't mark a cert as de-vs compliant if it leads to SHA-1 sigs.
Werner Koch [Thu, 28 Feb 2019 13:43:42 +0000 (14:43 +0100)]
sm: Don't mark a cert as de-vs compliant if it leads to SHA-1 sigs.

* sm/keylist.c (print_compliance_flags): Also check the digest_algo.
Add new arg 'cert'.
--

A certificate with algorithm sha1WithRSAEncryption can be de-vs
compliant (e.g. if the next in the chain used sha256WithRSAEncryption
to sign it and RSA is long enough) but flagging it as such is useless
because that certificate can't be used because it will create
signatures using the non-compliant SHA-1 algorithm.

Well, it could be used for encryption.  But also evaluating the
key-usage flags here would make it harder for the user to understand
why certain certificates are listed as de-vs compliant and others are
not.

Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit 2c75af9f65d15653ed1bc191f1098ae316607041)

Reworked to also pass the CERT.  Note that 2.2 won't get the PK
Screening feature.

Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit be69bf0cbd11cb8c0d452e07066669aacc6caafa)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name sm-Don-t-mark-a-cert-as-de-vs-compliant-if-it-leads-to-SH.patch

3 years agogpgsm: default to 3072-bit keys.
Daniel Kahn Gillmor [Thu, 7 Sep 2017 22:39:37 +0000 (18:39 -0400)]
gpgsm: default to 3072-bit keys.

* doc/gpgsm.texi, doc/howto-create-a-server-cert.texi: : update
default to 3072 bits.
* sm/certreqgen-ui.c (gpgsm_gencertreq_tty): update default to
3072 bits.
* sm/certreqgen.c (proc_parameters): update default to 3072 bits.
* sm/gpgsm.c (main): print correct default_pubkey_algo.

--

3072-bit RSA is widely considered to be 128-bit-equivalent security.
This is a sensible default in 2017.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
(cherry picked from commit 7955262151a5c755814dd23414e6804f79125355)
(cherry picked from commit 121286d9d1506dbaad9ba33bae2e459814fe5849)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name gpgsm-default-to-3072-bit-keys.patch

3 years agoagent: Fix for suggested Libgcrypt use.
Werner Koch [Fri, 22 Feb 2019 13:09:02 +0000 (14:09 +0100)]
agent: Fix for suggested Libgcrypt use.

* agent/divert-scd.c (divert_pkdecrypt): Skip a flags parameter.
--

The libgcrypt docs say that a "flags" parameter should always be used
in the input of pkdecrypt.  Thus we should allow that parameter also
when parsing an s-expression to figure out the algorithm for use with
scdaemon.

Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit a12c3a566e2e4b10bc02976a2819070877ee895c)
(cherry picked from commit 0a95b153811f36739d1b20f23920bad0bb07c68b)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name agent-Fix-for-suggested-Libgcrypt-use.patch

3 years agogpgscm: Build well even if NDEBUG defined.
NIIBE Yutaka [Mon, 25 Feb 2019 01:44:16 +0000 (10:44 +0900)]
gpgscm: Build well even if NDEBUG defined.

* gpgscm/scheme.c (gc_reservation_failure): Fix adding ";".
[!NDEBUG] (scheme_init_custom_alloc): Don't init seserved_lineno.

--

Cherry icked from master commit of:
e140c6d4f581be1a60a34b67b16430452f3987e8

In some build environment, NDEBUG is defined (although it's
bad practice).  This change supports such a situation.

GnuPG-bug-id: 3959
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
(cherry picked from commit 8161afb9dddaba839be92fbe9d85c05235eda825)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name gpgscm-Build-well-even-if-NDEBUG-defined.patch

3 years agogpg: Fix comparison.
Neal H. Walfield [Fri, 6 Oct 2017 09:51:39 +0000 (11:51 +0200)]
gpg: Fix comparison.

* g10/gpgcompose.c (literal_name): Complain if passed zero arguments,
not one or fewer.

Signed-off-by: Neal H. Walfield <neal@walfield.org>
(cherry picked from commit 1ed21eee79749b976b4a935f2279b162634e9c5e)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name gpg-Fix-comparison.patch

3 years agoscd: Distinguish cancel by user and protocol error.
NIIBE Yutaka [Wed, 20 Sep 2017 01:06:43 +0000 (10:06 +0900)]
scd: Distinguish cancel by user and protocol error.

* scd/apdu.h (SW_HOST_CANCELLED): New.
* scd/apdu.c (host_sw_string): Support SW_HOST_CANCELLED.
(pcsc_error_to_sw): Return SW_HOST_CANCELLED for PCSC_E_CANCELLED.
* scd/iso7816.c (map_sw): Return GPG_ERR_INV_RESPONSE for
SW_HOST_ABORTED and GPG_ERR_CANCELED for SW_HOST_CANCELLED.

Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
(cherry picked from commit 2396055c096884d521c26b76f26263a146207c24)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name scd-Distinguish-cancel-by-user-and-protocol-error.patch

3 years agocommon: Fix gnupg_wait_processes.
NIIBE Yutaka [Tue, 19 Sep 2017 03:28:43 +0000 (12:28 +0900)]
common: Fix gnupg_wait_processes.

* common/exechelp-posix.c (gnupg_wait_processes): Loop for r_exitcodes
even if we already see an error.

--

The value stored by waitpid for exit code is encoded;  It requires
decoded by WEXITSTATUS macro, regardless of an error.

For example, when one of processes is already exited and another is
still running, it resulted wrong value of in r_exitcodes[n].

Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
(cherry picked from commit eeb3da6eb717ed6a1a1069a7611eb37503e8672d)

Gbp-Pq: Topic from-2.2.14
Gbp-Pq: Name common-Fix-gnupg_wait_processes.patch

3 years agogpg: Emit an ERROR status if no key was found with --list-keys.
Werner Koch [Wed, 30 Jan 2019 13:40:26 +0000 (14:40 +0100)]
gpg: Emit an ERROR status if no key was found with --list-keys.

* g10/keylist.c (list_one): Emit status line.
--

Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit 140fda8c61422ec055c3f7e214cc35706c4320dd)
(cherry picked from commit 14ea581a1c040b53b0ad4c51136a7948363b1e4b)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name gpg-Emit-an-ERROR-status-if-no-key-was-found-with-list-ke.patch

3 years agoagent: Clear bogus pinentry cache, when it causes an error.
NIIBE Yutaka [Mon, 28 Jan 2019 03:58:13 +0000 (12:58 +0900)]
agent: Clear bogus pinentry cache, when it causes an error.

* agent/agent.h (PINENTRY_STATUS_*): Expose to public.
(struct pin_entry_info_s): Add status.
* agent/call-pinentry.c (agent_askpin): Clearing the ->status
before the loop, let the assuan_transact set ->status.  When
failure with PINENTRY_STATUS_PASSWORD_FROM_CACHE, it returns
soon.
* agent/findkey.c (unprotect): Clear the pinentry cache,
when it causes an error.

--

Cherry-picked from master commit of:
      02a2633a7f0b7d91aa48ea615fb3a0edfd6ed6bb

Debian-bug-id: 919856
GnuPG-bug-id: 4348
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
(cherry picked from commit 9109bb9919f84d5472b7e62e84b961414a79d3c2)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name agent-Clear-bogus-pinentry-cache-when-it-causes-an-error.patch

3 years agodirmngr: Fix initialization of assuan's nPth hook.
NIIBE Yutaka [Fri, 25 Jan 2019 01:15:39 +0000 (10:15 +0900)]
dirmngr: Fix initialization of assuan's nPth hook.

* dirmngr/dirmngr.c (main): Move assuan_set_system_hooks to...
(thread_init): ... here.

--

Cherry picked master commit of:
1f8817475f59ede3f28f57edc10ba56bbdd08b49

Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
(cherry picked from commit 7f4c3eb0a039621c564b6095ab5f810524843157)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name dirmngr-Fix-initialization-of-assuan-s-nPth-hook.patch

3 years agogpg: Allow generating Ed25519 key from an existing key.
Werner Koch [Wed, 30 Jan 2019 10:28:14 +0000 (11:28 +0100)]
gpg: Allow generating Ed25519 key from an existing key.

* g10/misc.c (map_pk_gcry_to_openpgp): Add EdDSA mapping.
--

Due to this missing mapping a "gpg --export --full-gen-key" with
selection "13 - Existing key" did not worked for an ed25519 key.

Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit 346a98fabe03adf2e202e36fc2aa24b1c2571154)
(cherry picked from commit 31d2a1eecaee766919b18bc42b918d9168f601f8)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name gpg-Allow-generating-Ed25519-key-from-an-existing-key.patch

3 years agodoc: Mark keyserver-options timeout and http-proxy as obsolete.
Werner Koch [Tue, 8 Jan 2019 10:21:07 +0000 (11:21 +0100)]
doc: Mark keyserver-options timeout and http-proxy as obsolete.

--

(cherry picked from commit 6c000d4b78b836686e5a2789cc88a41e465e4400)
(cherry picked from commit 9fd6ba268f1fdf77cc5baa6e8fd3ab28e432e49b)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name doc-Mark-keyserver-options-timeout-and-http-proxy-as-obso.patch

3 years agogpg: Stop early when trying to create a primary Elgamal key.
Werner Koch [Tue, 22 Jan 2019 09:06:15 +0000 (10:06 +0100)]
gpg: Stop early when trying to create a primary Elgamal key.

* g10/misc.c (openpgp_pk_test_algo2): Add extra check.
--

The problem is that --key-gen --batch with a parameter file didn't
detect that Elgamal is not capable of signing and so an error was only
triggered at the time the self-signature was created.  See the code
comment for details.

GnuPG-bug-id: 4329
Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit f97dc55ff1b041071bc3cbe98aa761bf77bb7ac8)
(cherry picked from commit f5d3b982e44c5cfc60e9936020102a598b635187)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name gpg-Stop-early-when-trying-to-create-a-primary-Elgamal-ke.patch

3 years agoSilence compiler warnings new with gcc 8.
Werner Koch [Mon, 17 Dec 2018 17:46:26 +0000 (18:46 +0100)]
Silence compiler warnings new with gcc 8.

* dirmngr/dns.c: Include gpgrt.h.  Silence -Warray-bounds also gcc.
* tests/gpgscm/scheme.c: Include gpgrt.h.
(Eval_Cycle): Ignore -Wimplicit-fallthrough.
--

The funny use of case and labels in the CASE macro seems confuse the
fallthrough detection.

Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit 21fc089148678f59edb02e0e16bed65b709fb972)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name Silence-compiler-warnings-new-with-gcc-8.patch

3 years agowks: Do not use compression for the encrypted data.
Werner Koch [Tue, 18 Dec 2018 07:21:03 +0000 (08:21 +0100)]
wks: Do not use compression for the encrypted data.

* tools/gpg-wks-client.c (encrypt_response): Add arg -z0.
* tools/gpg-wks-server.c (encrypt_stream): Ditto.
--

If for example a server was built without the development packages of
the compression libraries installed, the server will not be able to
decrypt a request.  In theory this can't happen due to the preference
system but it is just to easy to create the server's key using a
different version of gpg and then use gpg-wks-server built
differently.

For the short messages we exchange compression is not really required
and thus we better do without to make the system more robust.

Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit 70a8db0333e3c22403b3647f8b5f924f6dace719)
(cherry picked from commit 16424d8a34c7f6af1071fd19dfc180cb7d17c052)

Gbp-Pq: Topic from-2.2.13
Gbp-Pq: Name wks-Do-not-use-compression-for-the-encrypted-data.patch

3 years agofix spelling
Daniel Kahn Gillmor [Sun, 18 Nov 2018 22:33:55 +0000 (17:33 -0500)]
fix spelling

Gbp-Pq: Name fix-spelling.patch

3 years agoMake gpg-zip use tar from $PATH
Daniel Kahn Gillmor [Sun, 18 Nov 2018 22:29:52 +0000 (17:29 -0500)]
Make gpg-zip use tar from $PATH

Apparently there is no clean way to configure this from ./configure,
and upstream is deprecating gpg-zip anyway.  So just force-set tar to
be manually "tar" (meaning, that we should look in the $PATH at
runtime).

See also https://dev.gnupg.org/T4251 and https://bugs.debian.org/913582

Gbp-Pq: Name Make-gpg-zip-use-tar-from-PATH.patch

3 years agogpg: Prefer SHA-512 and SHA-384 in personal-digest-preferences.
Daniel Kahn Gillmor [Wed, 3 Jan 2018 17:34:26 +0000 (12:34 -0500)]
gpg: Prefer SHA-512 and SHA-384 in personal-digest-preferences.

* g10/keygen.c (keygen_set_std_prefs): prefer SHA-512
and SHA-384 by default.

--

In 8ede3ae29a39641a2f98ad9a4cf61ea99085a892, upstream changed the
defaults for --default-preference-list to advertise a preference for
SHA-512, without touching --personal-digest-preferences.  This makes
the same change for --personal-digest-preferences, since every modern
OpenPGP library supports them all.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic update-defaults
Gbp-Pq: Name gpg-Prefer-SHA-512-and-SHA-384-in-personal-digest.patch

3 years agogpg: Default to SHA-512 for all signature types on RSA keys.
Daniel Kahn Gillmor [Thu, 7 Sep 2017 22:49:35 +0000 (18:49 -0400)]
gpg: Default to SHA-512 for all signature types on RSA keys.

* g10/main.h (DEFAULT_DIGEST_ALGO): Use SHA512 instead of SHA256 in
--gnupg mode (leave strict RFC and PGP modes alone).
* configure.ac: Do not allow disabling sha512.
* g10/misc.c (map_md_openpgp_to_gcry): Always support SHA512.

--

SHA512 is more performant on most 64-bit platforms than SHA256, and
offers a better security margin.  It is also widely implemented.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic update-defaults
Gbp-Pq: Name gpg-Default-to-SHA-512-for-all-signature-types-on-RS.patch

3 years agoagent: Fix cancellation handling for scdaemon.
NIIBE Yutaka [Wed, 20 Sep 2017 01:42:28 +0000 (10:42 +0900)]
agent: Fix cancellation handling for scdaemon.

* agent/call-scd.c (cancel_inquire): Remove.
(agent_card_pksign, agent_card_pkdecrypt, agent_card_writekey)
(agent_card_scd): Don't call cancel_inquire.

--

Since libassuan 2.1.0, cancellation command "CAN" is handled within
the library, by assuan_transact.  So, cancel_inquire just caused
spurious "CAN" command to scdaemon which resulted an error.

Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
(cherry picked from commit 9f5e50e7c85aa8b847d38010241ed570ac114fc3)

Gbp-Pq: Topic from-master
Gbp-Pq: Name agent-Fix-cancellation-handling-for-scdaemon.patch

3 years agogpg: default to AES-256.
Daniel Kahn Gillmor [Thu, 7 Sep 2017 23:04:00 +0000 (19:04 -0400)]
gpg: default to AES-256.

* g10/main.h (DEFAULT_CIPHER_ALGO): Prefer AES256 by default.

--

It's 2017, and pretty much everyone has AES-256 available.  Symmetric
crypto is also rarely the bottleneck (asymmetric crypto is much more
expensive).  AES-256 provides some level of protection against
large-scale decryption efforts, and longer key lengths provide a hedge
against unforseen cryptanalysis.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
(cherry picked from commit 73ff075204df09db5248170a049f06498cdbb7aa)

Gbp-Pq: Topic from-master
Gbp-Pq: Name gpg-default-to-AES-256.patch

3 years agogpg: default to 3072-bit RSA keys.
Daniel Kahn Gillmor [Thu, 7 Sep 2017 22:41:10 +0000 (18:41 -0400)]
gpg: default to 3072-bit RSA keys.

* agent/command.c (hlp_genkey): update help text to suggest the use of
3072 bits.
* doc/wks.texi: Make example match default generation.
* g10/keygen.c (DEFAULT_STD_KEY_PARAM): update to
rsa3072/cert,sign+rsa3072/encr, and fix neighboring comment,
(gen_rsa, get_keysize_range): update default from 2048 to 3072).
* g10/keyid.c (pubkey_string): update comment so that first example
is the default 3072-bit RSA.

--

3072-bit RSA is widely considered to be 128-bit-equivalent security.
This is a sensible default in 2017.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
(cherry picked from commit 909fbca19678e6e36968607e8a2348381da39d8c)

Gbp-Pq: Topic from-master
Gbp-Pq: Name gpg-default-to-3072-bit-RSA-keys.patch

3 years agoagent: Avoid scheduled checks on socket when inotify is working.
Daniel Kahn Gillmor [Tue, 1 Nov 2016 04:57:44 +0000 (00:57 -0400)]
agent: Avoid scheduled checks on socket when inotify is working.

* agent/gpg-agent.c (handle_connections): When inotify is working, we
do not need to schedule a timer to evaluate whether we control our own
socket or not.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic gpg-agent-idling
Gbp-Pq: Name agent-Avoid-scheduled-checks-on-socket-when-inotify-.patch

3 years agoagent: Avoid tight timer tick when possible.
Daniel Kahn Gillmor [Tue, 1 Nov 2016 04:14:10 +0000 (00:14 -0400)]
agent: Avoid tight timer tick when possible.

* agent/gpg-agent.c (need_tick): Evaluate whether the short-phase
handle_tick() is needed.
(handle_connections): On each cycle of the select loop, adjust whether
we should call handle_tick() or not.
(start_connection_thread_ssh, do_start_connection_thread): Signal the
main loop when the child terminates.
* agent/call-scd.c (start_scd): Call interrupt_main_thread_loop() once
the scdaemon thread context has started up.

--

With this change, an idle gpg-agent that has no scdaemon running only
wakes up once a minute (to check_own_socket).

Thanks to Ian Jackson and NIIBE Yutaka who helped me improve some of
the blocking and corner cases.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic gpg-agent-idling
Gbp-Pq: Name agent-Avoid-tight-timer-tick-when-possible.patch

3 years agoagent: Allow threads to interrupt main select loop with SIGCONT.
Daniel Kahn Gillmor [Tue, 1 Nov 2016 04:45:23 +0000 (00:45 -0400)]
agent: Allow threads to interrupt main select loop with SIGCONT.

* agent/gpg-agent.c (interrupt_main_thread_loop): New function on
non-windows platforms, allows other threads to interrupt the main loop
if there's something that the main loop might be interested in.

--

For example, the main loop might be interested in changes in program
state that affect the timers it expects to see.

I don't know how to do this on Windows platforms, but i welcome any
proposed improvements.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic gpg-agent-idling
Gbp-Pq: Name agent-Allow-threads-to-interrupt-main-select-loop-wi.patch

3 years agoagent: Create framework of scheduled timers.
Daniel Kahn Gillmor [Tue, 1 Nov 2016 01:27:36 +0000 (21:27 -0400)]
agent: Create framework of scheduled timers.

agent/gpg-agent.c (handle_tick): Remove intermittent call to
check_own_socket.
(tv_is_set): Add inline helper function for readability.
(handle_connections) Create general table of pending scheduled
timeouts.

--

handle_tick() does fine-grained, rapid activity.  check_own_socket()
is supposed to happen at a different interval.

Mixing the two of them makes it a requirement that one interval be a
multiple of the other, which isn't ideal if there are different delay
strategies that we might want in the future.

Creating an extensible regular timer framework in handle_connections
should make it possible to have any number of cadenced timers fire
regularly, without requiring that they happen in cadences related to
each other.

It should also make it possible to dynamically change the cadence of
any regularly-scheduled timeout.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic gpg-agent-idling
Gbp-Pq: Name agent-Create-framework-of-scheduled-timers.patch

3 years agodirmngr: Avoid automatically checking upstream swdb.
Daniel Kahn Gillmor [Mon, 21 Nov 2016 04:09:24 +0000 (23:09 -0500)]
dirmngr: Avoid automatically checking upstream swdb.

* dirmngr/dirmngr.c (housekeeping_thread): Avoid automatically
checking upstream's software database.  In Debian, software updates
should be handled by the distro mechanism, and additional upstream
checks only confuse the user.
* doc/dirmngr.texi: document that --allow-version-check does nothing.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic dirmngr-idling
Gbp-Pq: Name dirmngr-Avoid-automatically-checking-upstream-swdb.patch

3 years agodirmngr: Avoid need for hkp housekeeping.
Daniel Kahn Gillmor [Sat, 29 Oct 2016 06:00:50 +0000 (02:00 -0400)]
dirmngr: Avoid need for hkp housekeeping.

* dirmngr/ks-engine-hkp.c (host_is_alive): New function.  Test whether
host is alive and resurrects it if it has been dead long enough.
(select_random_host, map_host, ks_hkp_mark_host): Use host_is_alive
instead of testing hostinfo_t->dead directly.
(ks_hkp_housekeeping): Remove function, no longer needed.
* dirmngr/dirmngr.c (housekeeping_thread): Remove call to
ks_hkp_housekeeping.

--

Rather than resurrecting hosts upon scheduled resurrection times, test
whether hosts should be resurrected as they're inspected for being
dead.  This removes the need for explicit housekeeping, and makes host
resurrections happen "just in time", rather than being clustered on
HOUSEKEEPING_INTERVAL seconds.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic dirmngr-idling
Gbp-Pq: Name dirmngr-Avoid-need-for-hkp-housekeeping.patch

3 years agodirmngr: hkp: Avoid potential race condition when some hosts die.
Daniel Kahn Gillmor [Sat, 29 Oct 2016 05:25:05 +0000 (01:25 -0400)]
dirmngr: hkp: Avoid potential race condition when some hosts die.

* dirmngr/ks-engine-hkp.c (select_random_host): Use atomic pass
through the host table instead of risking out-of-bounds write.

--

Multiple threads may write to hosttable[x]->dead while
select_random_host() is running.  For example, a housekeeping thread
might clear the ->dead bit on some entries, or another connection to
dirmngr might manually mark a host as alive.

If one or more hosts are resurrected between the two loops over a
given table in select_random_host(), then the allocation of tbl might
not be large enough, resulting in a write past the end of tbl on the
second loop.

This change collapses the two loops into a single loop to avoid this
discrepancy: each host's "dead" bit is now only checked once.

As Werner points out, this isn't currently strictly necessary, since
npth will not switch threads unless a blocking system call is made,
and no blocking system call is made in these two loops.

However, in a subsequent change in this series, we will call a
function in this loop, and that function may sometimes write(2), or
call other functions, which may themselves block.  Keeping this as a
single-pass loop avoids the need to keep track of what might block and
what might not.

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
Gbp-Pq: Topic dirmngr-idling
Gbp-Pq: Name dirmngr-hkp-Avoid-potential-race-condition-when-some.patch

3 years agoAvoid simple memory dumps via ptrace
Daniel Kahn Gillmor [Wed, 12 Aug 2015 00:28:26 +0000 (20:28 -0400)]
Avoid simple memory dumps via ptrace

This avoids needing to setgid gpg-agent.  It probably doesn't defend
against all possible attacks, but it defends against one specific (and
easy) one.  If there are other protections we should do them too.

This will make it slightly harder to debug the agent because the
normal user won't be able to attach gdb to it directly while it runs.

The remaining options for debugging are:

 * launch the agent from gdb directly
 * connect gdb to a running agent as the superuser

Upstream bug: https://dev.gnupg.org/T1211

Gbp-Pq: Topic block-ptrace-on-secret-daemons
Gbp-Pq: Name Avoid-simple-memory-dumps-via-ptrace.patch

3 years agoavoid regenerating defsincdate (use shipped file)
Daniel Kahn Gillmor [Mon, 29 Aug 2016 16:34:42 +0000 (12:34 -0400)]
avoid regenerating defsincdate (use shipped file)

upstream ships doc/defsincdate in its tarballs.  but doc/Makefile.am
tries to rewrite doc/defsincdate if it notices that any of the files
have been modified more recently, and it does so assuming that we're
running from a git repo.

However, we'd rather ship the documents cleanly without regenerating
defsincdate -- we don't have a git repo available (debian builds from
upstream tarballs) and any changes to the texinfo files (e.g. from
debian/patches/) might result in different dates on the files than we
expect after they're applied by dpkg or quilt or whatever, which makes
the datestamp unreproducible.

Gbp-Pq: Topic debian-packaging
Gbp-Pq: Name avoid-regenerating-defsincdate-use-shipped-file.patch

3 years agoavoid-beta-warning
Debian GnuPG Maintainers [Tue, 14 Apr 2015 14:02:31 +0000 (10:02 -0400)]
avoid-beta-warning

avoid self-describing as a beta

Using autoreconf against the source as distributed in tarball form
invariably results in a package that thinks it's a "beta" package,
which produces the "THIS IS A DEVELOPMENT VERSION" warning string.

since we use dh_autoreconf, i need this patch to avoid producing
builds that announce themselves as DEVELOPMENT VERSIONs.

See discussion at:

 http://lists.gnupg.org/pipermail/gnupg-devel/2014-November/029065.html

Gbp-Pq: Topic debian-packaging
Gbp-Pq: Name avoid-beta-warning.patch

3 years agognupg2 (2.2.12-1+deb10u2) buster-security; urgency=high
Daniel Kahn Gillmor [Fri, 1 Jul 2022 16:06:43 +0000 (17:06 +0100)]
gnupg2 (2.2.12-1+deb10u2) buster-security; urgency=high

  [ Roger Shimizu ]
  * d/control: Update Build-Depends: libgpg-error-dev (>= 1.35)

  [ Daniel Kahn Gillmor ]
  * fix broken status line (Closes: #1014157)

[dgit import unpatched gnupg2 2.2.12-1+deb10u2]

3 years agoImport gnupg2_2.2.12-1+deb10u2.debian.tar.xz
Daniel Kahn Gillmor [Fri, 1 Jul 2022 16:06:43 +0000 (17:06 +0100)]
Import gnupg2_2.2.12-1+deb10u2.debian.tar.xz

[dgit import tarball gnupg2 2.2.12-1+deb10u2 gnupg2_2.2.12-1+deb10u2.debian.tar.xz]

7 years agoImport gnupg2_2.2.12.orig.tar.bz2
Daniel Kahn Gillmor [Sat, 15 Dec 2018 01:17:16 +0000 (01:17 +0000)]
Import gnupg2_2.2.12.orig.tar.bz2

[dgit import orig gnupg2_2.2.12.orig.tar.bz2]