* agent/gpg-agent.c (create_server_socket): Also inhibit the removal
of the ssh socket if another agent process is already running.
GnuPG modern is not affected.
GnuPG-bug-id: 2258
Signed-off-by: Justus Winter <justus@g10code.com>
* agent/genkey.c (agent_ask_new_passphrase): Use a separate malloc for
PI2. Check return value of the malloc function.
* agent/command-ssh.c (ssh_identity_register): Use a separate malloc
for PI2. Wipe PI2.
--
For whatever stupid reasons I once allocated only one memory area and
split that into PI and PI2. This is actually a common pattern with
malloc but here we used a made up object size and do not take the
extra alignment required into account. One of these not yet hit by
a (sig)bus PC/VAX hacker bugs.
Instead of trying to fix the alignment, it is better to use a second
calloc for the second struct.
GnuPG-bug-id: 2112
Signed-off-by: Werner Koch <wk@gnupg.org>
Resolved conflicts:
agent/command-ssh.c
agent/genkey.c
(backport master commit ddf9dd135a)
* agent/call-pinentry.c (start_pinentry): Remove first instance of
sending the option.
--
That option to tell Pinentry about an external password cache was send
twice and the first one unconditionally.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/call-pinentry.c (start_pinentry): Add more strings.
--
We do this so that translations of these strings will be available at
the time a pinentry implements features which require these strings.
Signed-off-by: Werner Koch <wk@gnupg.org>
This is a backport from master
commit 02d5e12054
without the mode flag in the table.
* agent/command.c (cmd_get_passphrase): Read from the user cache.
--
We used to read the passphrase with mode CACHE_MODE_NORMAL but we put
it into the cache with CACHE_MODE_USER. However, agent_get_cache does
not yet distinguish between them and thus this does not change
anything.
(backported from commit 23d2ef83cd)
* agent/agent.h (agent_askpin): Add arguments keyinfo and cache_mode.
Update callers.
(agent_get_passphrase): Likewise.
(agent_clear_passphrase): New function.
(opt): Add field allow_external_cache.
* agent/call-pinentry.c (start_pinentry): Send "OPTION
allow-external-password-cache" to the pinentry.
(PINENTRY_STATUS_PASSWORD_FROM_CACHE): New constant.
(pinentry_status_cb): New function.
(agent_askpin): Add arguments keyinfo and cache_mode. If KEYINFO and
CACHE_MODE describe a cachable key, then send SETKEYINFO to the
pinentry. Pass PINENTRY_STATUS_CB to the "GETPIN" invocation. If the
passphrase was incorrect and PINENTRY_STATUS_PASSWORD_FROM_CACHE is
set, decrement PININFO->FAILED_TRIES.
(agent_get_passphrase): Add arguments keyinfo and cache_mode. If
KEYINFO and CACHE_MODE describe a cachable key, then send SETKEYINFO
to the pinentry.
(agent_clear_passphrase): New function.
* agent/call-pinentry.c (start_pinentry): Act upon new var,
allow_external_cache.
* agent/command.c (cmd_clear_passphrase): Call agent_clear_passphrase.
* agent/gpg-agent.c (oNoAllowExternalCache): New.
(opts): Add option --no-allow-external-cache.
(parse_rereadable_options): Set this option.
--
Signed-off-by: Neal H. Walfield <neal@g10code.com>
Based on commits:
3a9305439be201c20f25d7293cb31756b5c9f94fd3b5cad2342180845959
* agent/command-ssh.c (ssh_search_control_file): Check S before
upcasing it.
--
In contradiction to the comment we did not check the length of HEXGRIP
and thus the GPG_ERR_INV_LENGTH was never triggered.
Detected by Stack 0.3:
bug: anti-simplify
model: |
%cmp8 = icmp ne i32 %i.0, 40, !dbg !986
--> false
stack:
- /home/wk/s/gnupg/agent/command-ssh.c:1226:0
ncore: 2
core:
- /home/wk/s/gnupg/agent/command-ssh.c:1225:0
- buffer overflow
- /home/wk/s/gnupg/agent/command-ssh.c:1225:0
- buffer overflow
(backported from 2.1 commit 3529dd8bb5)
* sm/minip12.c: (p12_build) if jnlib_iconv_open fails, avoid
double-free of pwbuf.
--
Observed by Joshua Rogers <honey@internot.info>, who proposed a
slightly different fix.
Debian-Bug-Id: 773472
Added fix at a second place - wk.
* agent/protect-tool.c (opt_session_env): New.
(main): Pass session environment object to
gnupg_prepare_get_passphrase.
--
GnuPG-bug-id: 1402
The full story can be found at
https://bugzilla.redhat.com/show_bug.cgi?id=548528
Sorry for the delay.
* agent/gpg-agent.c (GCRY_THREAD_OPTION_PTH_IMPL): Do not use with
libgcrypt >= 1.6.
(main): Ditto.
* scd/scdaemon.c (GCRY_THREAD_OPTION_PTH_IMPL): Ditto.
(main): Ditto.
--
This is not anymore needed but kept for compatibility with Libgcrypt <
1.6.
--
GnuPG-bug-id: 1561
Note that this is not a complete solution. The libgpg-error include
directory has now a higher preference but ld may not pick up the right
library if another one is installed. The problem is that the -L
option and the -l options are not emitted separately by
gpg-error-config.
* agent/command-ssh.c (setup_ssh_env): Fix env setting.
--
gniibe reported this to gnupg-devel on 2012-07-04:
[...]
(2) UPDATESTARTUPTTY doesn't work to switch TTY for pinentry for
SSH.
[...]
Current implementation:
In the function start_command_handler_ssh, the logic puts
priority on ctrl->session_env which is initialized by
agent_init_default_ctrl. There are always GPG_TTY and TERM
defined, because lines around 968 in gpg-agent.c, it says:
/* Make sure that we have a default ttyname. */
While UPDATESTARTUPTTY updates opt.startup_env, it doesn't
affect at all.
Here is a patch to point the issue. Tested and works for me.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/command-ssh.c (struct control_file_s): Rename to
ssh_control_file_s.
(ssh_open_control_file, ssh_close_control_file)
(ssh_read_control_file, ssh_search_control_file): New.
(control_file_t): Rename and move to ...
* agent/agent.h (ssh_control_file_t): here.
* agent/command.c (do_one_keyinfo): Add args is_ssh, ttl, disabled,
and confirm. Rename unknown keytype indicator from '-' to 'X'. Extend
output.
(cmd_keyinfo): Add options --ssh-list and --with-ssh.
--
This extension allows the development of frontends to manage the
sshcontrol file.
Signed-off-by: Werner Koch <wk@gnupg.org>
(cherry picked from commit 50c98c7ed6)
Conflicts in agent/command.c (due to less information printed by
keyinfo) solved.
* agent/gpg-agent.c (opts, main): Add option --no-allow-mark-trusted.
Put this option into the gpgconf-list.
(main): Enable opt.allow_mark_trusted by default.
* tools/gpgconf-comp.c (gc_options_gpg_agent): Replace
allow-mark-trusted by no-allow-mark-trusted.
* agent/trustlist.c (agent_marktrusted): Always set the "relax" flag.
--
These changes have been in effect for the Gpg4win Windows version
since 2011-01-24 and thus first released with Gpg4win 2.1.0. Given
the current state of PKIX it does not make any sense to lure the Unix
user into false security by making it harder to trust self-signed or
CAcert certificates.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/gpg-agent.c [W32]: Include Several Windows header.
(opts): Change help text for enable-ssh-support.
(opts, main): Add option --enable-putty-support
(putty_support, PUTTY_IPC_MAGIC, PUTTY_IPC_MAXLEN): New for W32.
(agent_init_default_ctrl): Add and asssert call.
(putty_message_proc, putty_message_thread): New.
(handle_connections) [W32]: Start putty message thread.
* common/sysutils.c (w32_get_user_sid): New for W32 only
* tools/gpgconf-comp.c (gc_options_gpg_agent): Add
--enable-ssh-support and --enable-putty-support. Make the
configuration group visible at basic level.
* agent/command-ssh.c (serve_mmapped_ssh_request): New for W32 only.
--
This patch enables support for Putty. It has been tested with Putty
0.62 using an Unix created ssh key copied to the private-keys-v1.d
directory on Windows and with a manually crafted sshcontrol file. It
also works with a smartcard key.
May thanks to gniibe who implemented a proxy in Python to test the
putty/gpg-agent communication.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/command-ssh.c (file_to_buffer)
(ssh_handler_request_identities): Open streams in binary mode.
(start_command_handler_ssh): Factor some code out to ..
(setup_ssh_env): new function.
--
This is for now a theoretical fix because there is no ssh client yet
which uses the GnuPG style IPC. OpenSSL for Cygwin uses only a quite
similar one. gniibe suggested to implement that IPC style in
Libassuan so that a Cygwin version of OpenSSL may be used with GnuPG.
Signed-off-by: Werner Koch <wk@gnupg.org>
* agent/command-ssh.c (SPEC_FLAG_IS_ECDSA): New.
(struct ssh_key_type_spec): Add fields CURVE_NAME and HASH_ALGO.
(ssh_key_types): Add types ecdsa-sha2-nistp{256,384,521}.
(ssh_signature_encoder_t): Add arg spec and adjust all callers.
(ssh_signature_encoder_ecdsa): New.
(sexp_key_construct, sexp_key_extract, ssh_receive_key)
(ssh_convert_key_to_blob): Support ecdsa.
(ssh_identifier_from_curve_name): New.
(ssh_send_key_public): Retrieve and pass the curve_name.
(key_secret_to_public): Ditto.
(data_sign): Add arg SPEC and change callers to pass it.
(ssh_handler_sign_request): Get the hash algo from SPEC.
* common/ssh-utils.c (get_fingerprint): Support ecdsa.
* agent/protect.c (protect_info): Add flag ECC_HACK.
(agent_protect): Allow the use of the "curve" parameter.
* agent/t-protect.c (test_agent_protect): Add a test case for ecdsa.
* agent/command-ssh.c (ssh_key_grip): Print a better error code.
--
The 3 standard curves are now supported in gpg-agent's ssh-agent
protocol implementation. I tested this with all 3 curves and keys
generated by OpenSSH 5.9p1.
Using existing non-ssh generated keys will likely fail for now. To fix
this, the code should first undergo some more cleanup; then the fixes
are pretty straightforward. And yes, the data structures are way too
complicated.
(cherry picked from commit 649b31c663)
Solved conflicts:
agent/protect.c
agent/t-protect.c
common/ssh-utils.c (different variabale name)
* agent/command-ssh.c (ssh_signature_encoder_dsa): Rewrite.
--
Using es_fopenmem instead of a preallocated buffer is safer and easier
to read.
(cherry picked from commit f76a0312c3)
* agent/command-ssh.c: Remove dirent.h.
(control_file_s): Add struct item.
(rewind_control_file): New.
(search_control_file): Factor code out to ...
(read_control_file_item): New.
(ssh_handler_request_identities): Change to iterate over entries in
sshcontrol.
--
Formerly we scanned the private key directory for matches of entries
in sshcontrol. This patch changes it to scan the sshcontrol file and
thus considers only keys configured there. The rationale for this is
that it is common to have only a few ssh keys but many private keys.
Even if that assumption does not hold true, the scanning of the
sshcontrol file is faster than reading the directory and only then
scanning the ssh control for each directory entry.
(cherry picked from commit d2777f84be)
* agent/command-ssh.c (SSH_CONTROL_FILE_NAME): New macro to replace
the direct use of the string.
(struct control_file_s, control_file_t): New.
(open_control_file, close_control_file): New. Use them instead of
using fopen/fclose directly.
--
(cherry picked from commit 25fb53ab4a)
Fixed conflicts in some variabale names.
* agent/command-ssh.c (ssh_handler_request_identities): Do not call
card_key_available if the scdaemon is disabled.
--
(back ported from commit id 781e9746df)
* agent/command-ssh.c (open_control_file): Use_es_fopen to support
the "wx" mode flag.
--
This also patch also specifies a file mode parameter. However, this
will only be used with an updated version of es_stream which we have
not yet done.
--
This is required by newer mingw toolchain versions which demand that
winsock2.h is included before windows.h. Now, due to the use of
socket definitions in pth.h we need to include winsock2.h also in
pth.h, now pth.h is often included after an include of windows.h and
thus the compiler spits out a warning. To avoid that we include
winsock2.h at all places the compiler complains about.
* agent/call-pinentry.c (atfork_cb): Reset signal mask and signal
handler for child process.
(agent_popup_message_stop): Send SIGINT (was: SIGKILL).
--
pinentry-curses should be killed by SIGINT, so that it can reset
terminal settings, but it didn't work.
* findkey.c (agent_public_key_from_file): Fix use of
gcry_sexp_build_array.
--
A test case leading to a segv in Libgcrypt is
gpg-connect-agent \
"READKEY 9277C5875C8AFFCB727661C18BE4E0A0DEED9260" /bye
The keygrip was created by "monkeysphere s", which has a comment.
gcry_sexp_build_array expects pointers to the arguments which is quite
surprising. Probably ARG_NEXT was accidentally implemented wrongly.
Anyway, we can't do anything about it and thus need to fix the check
the users of this function.
Some-comments-by: Werner Koch <wk@gnupg.org>
agent/genkey.c: s/to to/to/
sm/*.c: s/failed to allocated/failed to allocate/
sm/certlist.c: s/should have not/should not have/
Consistency fix:
* g10/gpg.c, kbx/kbxutil.c, sm/gpgsm.c: uppercase after Syntax
Replace gcry_md_start_debug by gcry_md_debug in all files.
* agent/gpg-agent.c (fixed_gcry_pth_init): Use only if
GCRY_THREAD_OPTION_VERSION is 0
* scd/scdaemon.c (fixed_gcry_pth_init): Ditto.
--
Libgcrypt 1.6 will have some minor API changes. In particular some
deprecated macros and functions will be removed. PTH will also be
dropped in favor of a thread model neutral locking method.
* scripts/gitlog-to-changelog: New script. Taken from gnulib.
* scripts/git-log-fix: New file.
* scripts/git-log-footer: New file.
* scripts/git-hooks/commit-msg: New script.
* autogen.sh: Install commit-msg hook for git.
* doc/HACKING: Describe the ChangeLog policy.
* ChangeLog: New file.
* Makefile.am (EXTRA_DIST): Add new files.
(gen-ChangeLog): New.
(dist-hook): Run gen-ChangeLog.
Rename all ChangeLog files to ChangeLog-2011.
Fabian Keil found the reason for a SIGBUS:
In the "gpg-agent --daemon" case, main() calls pth_kill()
after the client has been forked, so when es_deinit() is
called on exit, acquiring the estream_list_lock seems to
cause pth to dereference a pointer located in a memory
region that has previously been free()'d.
My approach to fix it is different than his suggestion. It should
allow to continue all estream operations after a pth_kill except for
restarting pth.
Justus Winter