* kbx/keybox-search.c (blob_cmp_fpr_part): Reworked.
(has_short_kid, has_long_kid): Simplify.
--
The old code was too complicated and did not cope correctly a blob
having a mix of v5 and v4 keys.
Fixes-commit: 01329da8a7
GnuPG-bug-id: 5888
* agent/gpg-agent.c (main): The value of reliable_homedir_inotify
doesn't not related to nodetach, and it's only zero in the specific
condition.
* kbx/keyboxd.c (handle_connections): Remove the last argument.
(main): Remove reliable_homedir_inotify, as it's always one.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* agent/gpg-agent.c (reliable_homedir_inotify): Remove the global.
(handle_connections): Add reliable_homedir_inotify as an arg.
Don't call gnupg_inotify_watch_delete_self when it's not reliable.
(check_others_thread): No check of reliable_homedir_inotify repeatedly
in the loop.
* kbx/keyboxd.c (reliable_homedir_inotify): Remove the global.
(handle_connections): Add reliable_homedir_inotify as an arg.
(handle_tick): No check of reliable_homedir_inotify in the loop.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/keyboxd.c (create_directories): Following the behavior of
gpg-agent, call create_public_keys_directory after mkdir.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* g10/call-keyboxd.c (keydb_default_status_cb): New.
(keydb_update_keyblock): Add new status callback.
(keydb_insert_keyblock): Ditto.
(keydb_delete_keyblock): Ditto.
(search_status_cb): Also try the new status callback.
* sm/keydb.c (keydb_default_status_cb): New.
(keydb_insert_cert): Add new status callback.
(keydb_delete): Ditto
(search_status_cb): Also try the new status callback.
--
GnuPG-bug-id: 6838
* kbx/backend-sqlite.c (dblock_info_cb): New.
(create_or_open_database): Add arg ctrl. Add a 10 second timeout.
Avoid warning on error if not locked.
(be_sqlite_add_resource): Do not open the database here.
(be_sqlite_search): ... but do it here.
--
Note that we need to delay the initalization to the first use of the
database so that we actually have a recipient for the status messages.
GnuPG-bug-id: 6838
* kbx/kbx-client-util.c (prepare_data_pipe): Return an error.
(kbx_client_data_new): Recover from an error by use of D-lines.
(kbx_client_data_release): Handle the case of use of D-lines.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
This reverts commit 6944aefa3c.
--
The fd-passing works well on Windows with new libassuan (to be 3.0),
and it doesn't require ASSUAN_SOCKET_SERVER_FDPASSING actually.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/kbxserver.c (prepare_outstream): Use es_sysopen_nc
and avoid the use of translate_sys2libc_fd.
--
On Windows, it's better directly use the system HANDLE.
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/keybox-dump.c (_keybox_dump_find_dups): Close FP on the error
paths.
--
GnuPG-bug-id: 6495
Signed-off-by: zhangguangzhi <zhangguangzhi3@huawei.com>
* kbx/keybox-dump.c (_keybox_dump_find_dups): Close FP on the error
paths.
--
GnuPG-bug-id: 6495
Signed-off-by: zhangguangzhi <zhangguangzhi3@huawei.com>
* g10/call-keyboxd.c (gpg_keyboxd_deinit_session_data): Release
the assuan connection before kbx_client_data_release.
(open_context): Enable use of the data pipe.
* sm/keydb.c (gpgsm_keydb_deinit_session_data): Release the
assuan connection before kbx_client_data_release.
(open_context): Enable use of the data pipe.
* kbx/kbx-client-util.c (struct kbx_client_data_s): Add THD field.
(prepare_data_pipe): Close the pipe output end as it's been sent
already. Remember the KCD->THD, so that it can be joined later.
(datastream_thread): Finish when reading no data from the pipe.
(kbx_client_data_release): Join the thread. Then, we can safely
call es_fclose on the FP.
--
GnuPG-bug-id: 6512
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/kbx-client-util.c (prepare_data_pipe): Use _get_osfhandle
for pipe to be used for sentfd.
[HAVE_W32_SYSTEM] (datastream_thread): Add the case of NREAD==0.
* tools/gpg-connect-agent.c (do_sendfd): Use es_syshd instead
of es_fileno.
[HAVE_W32_SYSTEM] (do_open): Use %p for formating HANDLE.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/kbxserver.c (kbxd_start_command_handler): No fd-passing udner
Windows.
--
file descriptor passing does not work reliable in libassuan for
Windows and we actually don't need it here. It is not even used by
gpg or gpgsm. As soon as we enable fd-passing in gpgme for Windows
and see that it is robust enough we should back out this patch.
* kbx/keybox-init.c (DEFAULT_LL_BUFFER_SIZE): New.
--
A simple gpg --check-sigs benchmark showed on Linux a small
performance peak at around 64k (5m52 vs. 6m8 for 128k and 6m33 for
system size).
* kbx/keybox-init.c (ll_buffer_size): New var intialized to 128k
(stream_buffers): New var.
(keybox_set_buffersize): New.
(_keybox_ll_open, _keybox_ll_close): Implement buffering.
* sm/gpgsm.c (oKbxBufferSize): New.
(opts): Add option --kbx-buffer-size.
(main): Call keybox_set_buffersize.
* g10/gpg.c: Include keybox.h.
* (oKbxBufferSize): New.
(opts): Add option --kbx-buffer-size.
(main): Call keybox_set_buffersize.
--
Running a test on Windows using a pubring.kbx with
Total number of blobs: 2098
openpgp: 1294
x509: 803
and a size of 42MiB with
gpgsm -k --with-validation --disable-dirmngr --kbx-buffer-size N >nul
gives these performance figures using procmon
| N(k) | file events | time(s) |
|------+-------------+---------|
| 0 | 4900000 | 86 |
| 16 | 2456000 | 58 |
| 32 | 1233000 | 43 |
| 64 | 622000 | 37 |
| 128 | 317000 | 32 |
| 256 | 164000 | 31 |
| 512 | 88000 | 30 |
Using _open instead of CreateFile give the same number of file events
but increased the time slight by one second for the measured buffer
size of 64k and 128k. Benchmarks for gpg have not been conducted.
* kbx/keybox-defs.h (KEYBOX_LL_OPEN_READ)
(KEYBOX_LL_OPEN_UPDATE, KEYBOX_LL_OPEN_CREATE): New.
* kbx/keybox-init.c (_keybox_ll_open): New. Replace all keybox use of
es_fopen by this function.
(_keybox_ll_close): New. Replace all keybox use of es_fclose by this
function.
--
Note that this has not been done for the utilities and the backend-kbx
of keyboxd.
* kbx/keybox-search.c (open_file): Use sysopen and sequential.
* kbx/keybox-update.c (create_tmp_file): Ditto.
(blob_filecopy): Ditto.
(keybox_set_flags): Ditto.
(keybox_delete): Ditto.
(keybox_compress): Ditto.
--
Under Windows "sysopen" requests that direct API calls (CreateFile et
al.) are used instead of the libc wrappers. This may or may not
improve the performance.
Using "sequential" is a hint to Windows to assume that a file is in
general access in a sequential manner. This will have an affect only
with a future libgpg-error.
* kbx/keyboxd.h (kbxd_status_printf): New.
* kbx/backend-support.c (be_return_pubkey): Use kbxd_status_printf.
* kbx/kbxserver.c (kbxd_status_printf): New.
* kbxd_start_command_handler (kbxd_start_command_handler): Don't
use set_assuan_context_func, because the function pointer is
shared by multiple threads.
--
GnuPG-bug-id: 5948
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/keybox-search.c (blob_cmp_fpr_part): Don't change FPROFF, since
it's caller which tweaks the offset.
(has_short_kid, has_long_kid): Examine the key flags to determine if
fingerprint 32 or 20.
--
GnuPG-bug-id: 5888
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/keyboxd.c (oStealSocket): New const.
(opts): Add option.
(steal_socket): New file global flag.
(main): Set option.
(create_server_socket): Implement option.
--
* kbx/backend-kbx.c (be_kbx_search): Initialize skipped_long_blobs
value which is passed to keybox_search and incremented there.
(be_kbx_seek): Likewise.
--
GnuPG-bug-id: 5393
Co-authored-by: NIIBE Yutaka <gniibe@fsij.org>
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
* kbx/keybox-blob.c (create_blob_header): Fix the flag to match no
UBID entry,
(create_blob_finish): Fix the length of data to be hashed.
--
GnuPG-bug-id: 5573
Fixes-commit: 915297705a
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* kbx/keybox-search-desc.h (struct keydb_search_desc): New flag
name_used.
* common/userids.c (classify_user_id): Set flag.
* kbx/kbxserver.c (struct search_backing_store_s): New.
(cmd_search): use a backing store for the const pointers.
(kbxd_start_command_handler): Release the backing store.
--
Well, the search object partly uses buffers but also const
pointers (for strings and the serial number). This when assigning
such objects to an another one we should really take a deep copy and
not just copy the pointer. The more clean solution would have been to
provide a storage option the search object but that needs checking the
code at too many places so that I decided to use a separate backing
store array here.
Signed-off-by: Werner Koch <wk@gnupg.org>
Werner Koch