Andre Noll [Thu, 7 Jan 2021 23:55:13 +0000 (00:55 +0100)]
alsa_write: Avoid hung tasks.
Without this fix the alsa writer may never shutdown, occupying a
slot of audiod. The hang was observed with in conjunction to the
udp receiver and the prebuffer filter. See also the recent commit 23913cbbfc98.
Andre Noll [Fri, 16 Oct 2020 13:49:47 +0000 (15:49 +0200)]
aft: Avoid NULL pointer dereference.
osl_get_object() must not be called with a NULL row pointer. Currently
this may happen on blob events. This patch avoids the problem and
makes sure we catch this programming error early.
Andre Noll [Fri, 1 Jan 2021 15:05:56 +0000 (16:05 +0100)]
prebuffer: Bail out on bufffer tree errors.
Without this, the prebuffer filter might stay active forever, occupying
a slot of para_audiod until no more slots are available. This was
observed with the udp receiver.
Andre Noll [Mon, 1 Jun 2020 17:28:01 +0000 (19:28 +0200)]
Replace /* fallthrouth*/ by __attribute__ ((fallthrough));
Some (recent) gcc versions still complain because the switch case
falls through. The new code causes a warning on old compilers which
don't know the fallthrough attribute, but this is still better than
seeing the warning on systems with a recent compiler.
Andre Noll [Fri, 29 May 2020 23:22:49 +0000 (01:22 +0200)]
aft: Avoid invalid read.
A short chunk table is fatal for all audio formats except aac,
which employs dynamic chunks. The below valgrind spat was found when
para_server tried to open an aac audio file. Setting afhi->chunk_table
to NULL in this case should fix it.
==17667== Invalid read of size 4
==17667== at 0x805A862: write_u32 (portable_io.h:95)
==17667== by 0x805A862: save_chunk_table (aft.c:402)
==17667== by 0x805A862: save_chunk_table (aft.c:395)
==17667== by 0x805DDE6: save_afd (aft.c:616)
==17667== by 0x805DDE6: open_and_update_audio_file (aft.c:1113)
==17667== by 0x8058AA2: open_next_audio_file (afs.c:425)
==17667== by 0x8058AA2: execute_server_command (afs.c:867)
==17667== by 0x8058AA2: command_post_select.part.0 (afs.c:921)
==17667== by 0x8063062: call_post_select (sched.c:80)
==17667== by 0x8063062: sched_post_select (sched.c:106)
==17667== by 0x8063062: schedule (sched.c:159)
==17667== by 0x8059643: afs_init (afs.c:1006)
==17667== by 0x804D747: init_afs (server.c:529)
==17667== by 0x804D747: server_init (server.c:601)
==17667== by 0x804D747: main (server.c:690)
==17667== Address 0x4d7dcd0 is 0 bytes after a block of size 40 alloc'd
==17667== at 0x40365E2: malloc (vg_replace_malloc.c:309)
==17667== by 0x8053AB6: para_malloc (string.c:63)
==17667== by 0x805B20D: load_chunk_table (aft.c:415)
==17667== by 0x805DD65: open_and_update_audio_file (aft.c:1103)
==17667== by 0x8058AA2: open_next_audio_file (afs.c:425)
==17667== by 0x8058AA2: execute_server_command (afs.c:867)
==17667== by 0x8058AA2: command_post_select.part.0 (afs.c:921)
==17667== by 0x8063062: call_post_select (sched.c:80)
==17667== by 0x8063062: sched_post_select (sched.c:106)
==17667== by 0x8063062: schedule (sched.c:159)
==17667== by 0x8059643: afs_init (afs.c:1006)
==17667== by 0x804D747: init_afs (server.c:529)
==17667== by 0x804D747: server_init (server.c:601)
==17667== by 0x804D747: main (server.c:690)
Andre Noll [Mon, 16 Mar 2020 17:44:07 +0000 (18:44 +0100)]
mp: Always set mp_context to NULL on errors.
In mp_init(), if mp_yyparse() fails, we return early without setting
the result pointer to NULL. This does not matter much because both
callers of mood.c pass in NULL, but still..
Andre Noll [Fri, 12 Jun 2020 01:34:32 +0000 (03:34 +0200)]
vss: Rework fec client setup.
The current fec code assumes that the chunks of the audio file form a
contigous buffer. At least for aac/m4a this is not true, which is
why streaming m4a files over udp never worked well.
This patch should be a big improvement in this regard. We now copy
the chunks to preallocated buffers, which also makes the code easier
to follow because we can get rid of the two extra buffers in struct
fec_client.
Andre Noll [Sat, 6 Jun 2020 16:50:48 +0000 (18:50 +0200)]
gui: Adjust position of num_played value.
If the terminal window is only 80 characters wide, the num_played
value may be adjacent to the bitrate value, with no space between.
This happens only if the num_played value is bigger than 99, and only
with the "colorful blackness" theme.
Andre Noll [Thu, 11 Jun 2020 13:28:49 +0000 (15:28 +0200)]
Avoid audiod hangs when prebuffer filter is active.
This filter misses to honor task notifications. As a result of this
omission, when CTRL+C is pressed while audiod is running in foreground
mode and the prebuffer filter is active, the audiod process hangs.
Andre Noll [Mon, 23 Mar 2020 20:19:14 +0000 (21:19 +0100)]
Merge branch 'refs/heads/t/stale-pointer-fix'
This bug only triggered if the kernel changes the address of the memory
mapping of the audio file table after a file was added, and a subsequent
operation would access the then stale pointer.
Cooking for a week.
* refs/heads/t/stale-pointer-fix:
Don't use strdup() to copy hash.
aft: Avoid stale pointer pointer reference.
Andre Noll [Sun, 2 Feb 2020 15:26:16 +0000 (16:26 +0100)]
aft: Avoid stale pointer pointer reference.
We can't rely on the ->hash and ->path pointers of struct
status_item_ls_data because they both become stale when the audio
file table gets remapped. This happens for example when a new audio
file is added while another audio file is currently open. A subsequent
addblob command then triggers a re-initialization of the status items,
which reads from the stale pointer location.
This usually results in garbage in the stat output, but can also lead
to a segfault or worse. Avoid this from happening by creating a copy
of the hash and the path.
This bug has been present for a long time. It was hard to debug
because often no invalid memory accesses occur, hence valgrind does
not complain.
Andre Noll [Sun, 15 Mar 2020 14:51:20 +0000 (15:51 +0100)]
Merge branch 'refs/heads/t/mixer'
A single patch for para_mixer which makes the subcommands which
sleep a bit more robust.
The merge conflicted due to commit d6b25bf854c1 (mixer: fade: Handle
empty mood strings gracefully) from half a year ago, but the resolution
was obvious.
Cooking for nine months.
* refs/heads/t/mixer:
mixer: sleep/snooze: Close mixer during sleep.
Andre Noll [Sun, 24 Nov 2019 16:33:51 +0000 (17:33 +0100)]
mp3_afh: Don't bail out on short files.
It's perfectly OK to have mp3 files which are shorter than two seconds.
For example lexico.com offers mp3 downloads to learn the pronuciation
of single words. These files are often shorter than one second.
The check for short length files predates the git history. Most likely
it was added for no good reason, so drop it.
Andre Noll [Wed, 11 Sep 2019 20:12:32 +0000 (22:12 +0200)]
openssl: Fix (harmless) memory leaks at exit.
Without this, valgrind --leak-check=full --show-leak-kinds=all
complains about four memory blocks reachable at exit, see below.
The first leak is fixed by the new call to ERR_remove_thread_state(),
which frees the error queue of the current thread. To squash the
other three, the call to EVP_cleanup() is needed to remove all of
openssl's internal ciphers and digests.
To make this work on both openssl-1.0 and for openssl-1.1, we need
to add additional header checks for configure.
---
==2818== 12 bytes in 1 blocks are still reachable in loss record 1 of 4
==2818== at 0x40355EE: malloc (vg_replace_malloc.c:309)
==2818== by 0x41B887C: CRYPTO_malloc (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x42459B7: lh_insert (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4248542: int_thread_set_item (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4249920: ERR_get_state (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x424A255: ERR_put_error (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4222EF6: RSA_padding_check_PKCS1_OAEP_mgf1 (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x422306A: RSA_padding_check_PKCS1_OAEP (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x421F921: RSA_eay_private_decrypt (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x804F2E9: apc_priv_decrypt (openssl.c:206)
==2818== by 0x804DB59: client_post_select (client_common.c:321)
==2818== by 0x804CC57: call_post_select (sched.c:80)
==2818== by 0x804CC57: sched_post_select (sched.c:106)
==2818== by 0x804CC57: schedule (sched.c:159)
==2818==
==2818== 64 bytes in 1 blocks are still reachable in loss record 2 of 4
==2818== at 0x40355EE: malloc (vg_replace_malloc.c:309)
==2818== by 0x41B887C: CRYPTO_malloc (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4245643: lh_new (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x424811F: int_thread_get (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4248511: int_thread_set_item (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4249920: ERR_get_state (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x424A255: ERR_put_error (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4222EF6: RSA_padding_check_PKCS1_OAEP_mgf1 (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x422306A: RSA_padding_check_PKCS1_OAEP (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x421F921: RSA_eay_private_decrypt (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x804F2E9: apc_priv_decrypt (openssl.c:206)
==2818== by 0x804DB59: client_post_select (client_common.c:321)
==2818==
==2818== 96 bytes in 1 blocks are still reachable in loss record 3 of 4
==2818== at 0x40355EE: malloc (vg_replace_malloc.c:309)
==2818== by 0x41B887C: CRYPTO_malloc (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4245628: lh_new (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x424811F: int_thread_get (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4248511: int_thread_set_item (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4249920: ERR_get_state (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x424A255: ERR_put_error (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4222EF6: RSA_padding_check_PKCS1_OAEP_mgf1 (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x422306A: RSA_padding_check_PKCS1_OAEP (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x421F921: RSA_eay_private_decrypt (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x804F2E9: apc_priv_decrypt (openssl.c:206)
==2818== by 0x804DB59: client_post_select (client_common.c:321)
==2818==
==2818== 400 bytes in 1 blocks are still reachable in loss record 4 of 4
==2818== at 0x40355EE: malloc (vg_replace_malloc.c:309)
==2818== by 0x41B887C: CRYPTO_malloc (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x424988C: ERR_get_state (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x424A255: ERR_put_error (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x4222EF6: RSA_padding_check_PKCS1_OAEP_mgf1 (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x422306A: RSA_padding_check_PKCS1_OAEP (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x421F921: RSA_eay_private_decrypt (in /usr/local/lib/libcrypto.so.1.0.0)
==2818== by 0x804F2E9: apc_priv_decrypt (openssl.c:206)
==2818== by 0x804DB59: client_post_select (client_common.c:321)
==2818== by 0x804CC57: call_post_select (sched.c:80)
==2818== by 0x804CC57: sched_post_select (sched.c:106)
==2818== by 0x804CC57: schedule (sched.c:159)
==2818== by 0x804AB2D: main (client.c:656)
Andre Noll [Fri, 28 Feb 2020 17:13:50 +0000 (18:13 +0100)]
Merge branch 'refs/heads/t/ogg_afh'
A patch which teaches the ogg/* audio format handlers to report the
correct length for files with holes. A bug was found in that patch
after the branch had been merged to next, so there's a fixup commit
on top.
The second patch in this series fixes an issue with
ogg_page_granulepos() that can result in incorrect estimates for the
duration of files that use the ogg container format.
Cooking for a year.
* refs/heads/t/ogg_afh:
ogg_afh_common.c: Check return value of ogg_page_granulepos().
ogg_afh_common: Fix signedness issue.
ogg: Detect missing ogg pages.
Andre Noll [Sat, 2 Nov 2019 13:47:15 +0000 (14:47 +0100)]
vss: Improve error diagnostics.
Unless in debug loglevel, para_server prints
Nov 01 17:06:03 (4) (18596) open_next_audio_file: key not found in rbtree
Nov 01 17:06:03 (4) (18595) recv_afs_result: did not receive open fd from afs
which does not include information about what was actually received.
Andre Noll [Sat, 2 Nov 2019 13:59:29 +0000 (14:59 +0100)]
command.c: Also invalidate play time if no audio file is open.
Without this, the time string of the previous audio file is continued
to be shown in the top window of para_gui when stopped. It gets
wiped out eventually, but this can take a few seconds. With the patch
applied, the time string disappears immediately.
Andre Noll [Sat, 23 Nov 2019 17:20:21 +0000 (18:20 +0100)]
Merge branch 'refs/heads/t/ssh'
A medium sized series which teaches both crypto backends about the
RFC4716 key format, which has become the default for ssh-keygen as
of openssh-7.8.
Was cooking for almost a year.
* refs/heads/t/ssh:
manual: Instruct the user to create RFC4716 keys.
openssl: Add support for RFC4716 keys
openssl: Move get_private_key() down.
gcrypt: Add support for RFC4716 private keys.
crypt: Rename decoding functions.
gcrypt: Introduce read_openssh_bignum().
gcrypt: Factor out read_pem_rsa_params().
gcrypt: Let read_bignum() return bits, not bytes.
gcrypt: Let decode_key() return blob size through additional argument.
gcrypt: Drop unnecessary arguments of decode_key().
Andre Noll [Tue, 20 Aug 2019 07:10:53 +0000 (09:10 +0200)]
Avoid warning about sys/sysctl.h on glibc-2.30.
From glibc-2.30 NEWS:
The Linux-specific <sys/sysctl.h> header and the sysctl function have
been deprecated and will be removed from a future version of glibc.
Compilation against the glibc-2.30 headers results in the following warning:
In file included from ipc.c:10:
/usr/include/sys/sysctl.h:21:2: warning: #warning "The <sys/sysctl.h> header is deprecated and will be removed." [-Wcpp]
On NetBSD and FreeBSD, however, we still need to include the header
to get the declaration of sysctlbyname(3).
This patch changes ipc.c to include sys/sysctl.h only if __FreeBSD__ or
__NetBSD__ is defined. Also remove the pointless check for __APPLE__.
Andre Noll [Sun, 1 Sep 2019 11:18:41 +0000 (13:18 +0200)]
Merge branch 'refs/heads/t/afh'
A couple of patches which remove ->init of struct audio_format_handler
and struct receiver.
Cooking for 9 months.
* refs/heads/t/afh:
Remove ->init() of struct receiver.
afh: Constify definition of audio format handlers.
afh: Introduce audio_format_names[].
afh: Get rid of dummy entry at the end of afl[].
afh: Minor simplification for afh_get_chunk().
afh: Move audio_format_name() up.
Andre Noll [Sat, 24 Aug 2019 11:38:29 +0000 (13:38 +0200)]
Merge branch 'refs/heads/t/compress'
A short series which overhauls the algorithm behind the compress
filter and its documentation.
Cooking for almost a year.
* refs/heads/t/compress:
compress: Overhaul the meaning of --aggressiveness.
compress: Apply damping later.
compress: Warn when samples are clipped.
compress: Document and sanity-check command line options.
compress: Fix off by one in help of --target-level.
Andre Noll [Thu, 4 Jul 2019 07:04:56 +0000 (09:04 +0200)]
ogg_afh_common.c: Check return value of ogg_page_granulepos().
This function may return -1, and we do not want to use a negative
granule position for the computation of the duration of the ogg file.
Note that oac_get_file_info() loops over all ogg pages twice, but
the second loop does not have the same problem as we already check
the return value there.
Andre Noll [Thu, 6 Jun 2019 06:28:58 +0000 (08:28 +0200)]
ogg_afh_common: Fix signedness issue.
The previous patch introduced a regression because "granule" was made
unsigned by mistake but needs to be a signed variable. This can cause
the loop in oac_get_file_info() to not terminate which eventually
results in an abort due to an allocation failure:
Andre Noll [Sat, 23 Mar 2019 15:51:56 +0000 (16:51 +0100)]
mixer: sleep/snooze: Close mixer during sleep.
This makes these subcommands work in case the audio device goes away
temporarily during the sleep interval. This does not work at the moment
because we connect to the mixer only once at startup. The mixer handle
becomes invalid when the audio device goes away, so fade-in fails.
Another drawback of this "connect only once" design is that the
mixer device is opened even for the help subcommand, which is not
only silly but also has the effect that the help subcommand fails if
the mixer device can not be opened.
To improve on this, we have the individual subcommand handlers open and
close the mixer device. Consequently, the command handlers no longer
take a struct mixer_handle * argument. To obtain a mixer handle,
command handlers must now call the new open_mixer_and_set_channel()
at startup, and m->close() on return to free the handle.
Andre Noll [Wed, 1 Aug 2018 12:13:07 +0000 (14:13 +0200)]
play: New option: --end-of-playlist.
If the end of the playlist is reached, para_play starts over with the
first file of the playlist. This commit provides the --end-of-playlist
option to control the behaviour. The new option has three pre-defined
values: stop, loop, quit with "loop" being the default. Since lopsub
features multi-valued options, the implementation is pretty simple.
Andre Noll [Sun, 27 Jan 2019 23:00:40 +0000 (00:00 +0100)]
aft: Avoid invalid read.
This happens with dynamic chunks. In this case we don't reserve any
space for the chunk table, yet the code copies one entry.
==6827== Invalid read of size 4
==6827== at 0x805A7C6: write_u32 (portable_io.h:95)
==6827== by 0x805A7C6: save_chunk_table (aft.c:402)
==6827== by 0x805E035: save_afd (aft.c:616)
==6827== by 0x805E035: open_and_update_audio_file (aft.c:1097)
==6827== by 0x80589B5: open_next_audio_file (afs.c:425)
==6827== by 0x80589B5: execute_server_command (afs.c:867)
==6827== by 0x80589B5: command_post_select.part.3 (afs.c:921)
==6827== by 0x80632CE: call_post_select (sched.c:80)
==6827== by 0x80632CE: sched_post_select (sched.c:106)
==6827== by 0x80632CE: schedule (sched.c:159)
==6827== by 0x8059555: afs_init (afs.c:1006)
==6827== by 0x804D53A: init_afs (server.c:528)
==6827== by 0x804D53A: server_init (server.c:600)
==6827== by 0x804D53A: main (server.c:689)
==6827== Address 0x58c4a20 is 0 bytes after a block of size 40 alloc'd
==6827== at 0x40345EE: malloc (vg_replace_malloc.c:299)
==6827== by 0x80539C5: para_malloc (string.c:63)
==6827== by 0x805B38A: load_chunk_table (aft.c:415)
==6827== by 0x805DFC4: open_and_update_audio_file (aft.c:1087)
==6827== by 0x80589B5: open_next_audio_file (afs.c:425)
==6827== by 0x80589B5: execute_server_command (afs.c:867)
==6827== by 0x80589B5: command_post_select.part.3 (afs.c:921)
==6827== by 0x80632CE: call_post_select (sched.c:80)
==6827== by 0x80632CE: sched_post_select (sched.c:106)
==6827== by 0x80632CE: schedule (sched.c:159)
==6827== by 0x8059555: afs_init (afs.c:1006)
==6827== by 0x804D53A: init_afs (server.c:528)
==6827== by 0x804D53A: server_init (server.c:600)
==6827== by 0x804D53A: main (server.c:689)
Andre Noll [Fri, 21 Dec 2018 17:11:40 +0000 (18:11 +0100)]
build: Check also for CRYPTO_cleanup_all_ex_data.
FreeBSD 12 does not seem to have it. This is not fatal since omitting
the call to CRYPTO_cleanup_all_ex_data() just results in a harmless
memory leak at exit.
Andre Noll [Fri, 21 Dec 2018 15:43:54 +0000 (16:43 +0100)]
build: Detect openssl library/header mismatch.
This catches the case where openssl-1.0 headers are used for compiling
but openssl-1.1 libraries for linking. Without the check that is added
by this commit the compilation succeeds in this case but the executable
segfaults on the attempt to modify the RSA structure directly. With
the new check, configure fails gracefully.
This happened on a FreeBSD system where the openssl-1.0 headers were
installed in /usr/local and openssl-1.1 headers in /usr.
Andre Noll [Sun, 30 Sep 2018 21:39:58 +0000 (23:39 +0200)]
mp3_afh: Move mode_text[] into header_mode().
The array is only used in header_mode(), so it should be local to
the function. Also, it's confusing to let the last element play the
role of the invalid header mode, and it's clearer to use ARRAY_SIZE()
instead of spelling out the index of the last element. Next, it's
unnecessary to write to the mp3header structure, so avoid this and
mark the pointer argument const. Finally, the patch makes the array
constant, not only the mode strings.
Andre Noll [Sat, 22 Dec 2018 13:54:24 +0000 (14:54 +0100)]
build: Remove hard-coded include path.
Compilers generally have /usr/local/include in their default search
path. And if they don't, that has to be considered a local policy
which should not be ignored. The user can always override it by
setting CPPFLAGS.
Andre Noll [Sat, 22 Dec 2018 13:00:14 +0000 (14:00 +0100)]
build: Fix CPPFLAGS for openssl.
The effect of the wrong file name was that --with-openssl-headers
was effectively ignored.
This bug was introduced one year ago in commit a4c2c4f9c7cd (crypt:
Rename RSA functions) which renamed the crypt.c file to openssl.c but
missed to adjust the Makefile accordingly.
Andre Noll [Sun, 30 Sep 2018 20:58:25 +0000 (22:58 +0200)]
vss: Add documentation of MAP_POPULATE.
It should be clear why the vss should ask the kernel to prefault the
page tables for the mapping, but it's less clear why we might need
to define the preprocessor symbol.
Andre Noll [Fri, 31 Aug 2018 10:59:06 +0000 (12:59 +0200)]
manual: Instruct the user to create RFC4716 keys.
We support both types of keys, but it would be lovely to get rid of
the open-coded parser for PEM keys in grcypt.c. This is the first
step to deprecate the use of PEM keys.
Andre Noll [Thu, 30 Aug 2018 13:32:34 +0000 (15:32 +0200)]
openssl: Add support for RFC4716 keys
The decode_private_key() helper of gcrypt.c base64-decodes the key
but does not read the RSA bignums. It is thus independent of gcrypt
and openssl. To add support for RFC4716 keys to the openssl backend,
we need to move it to crypt_common.c, making it a crypo backend
function. That is, non-static and declared in crypt_backend.h. The
patch also documents the function using doxgen comments.
With decode_private_key() and find_openssh_bignum_offset() to
our disposal, supporting RFC4716 keys is simple. We only need to
initialize the openssl-specific rsa structure with the six bignums
stored in a private RFC4716 key. This is implemented in the new
read_private_rsa_params() which calls the existing read_bignum()
six times.
The fields of the rsa structure are exposed to applications in
openssl-1.0, but the structure was made opaque in openssl-1.1. We
use the existing HAVE_RSA_SET0_KEY macro to decide whether or not
we must use the accessor functions of openssl-1.1 to initialize the
rsa structure.
Andre Noll [Tue, 28 Aug 2018 18:38:52 +0000 (20:38 +0200)]
gcrypt: Add support for RFC4716 private keys.
This teaches the gcrypt backend to parse private keys generated
with ssh-keygen -m RFC4716. Support for the openssl backend will be
implemented in a subsequent patch. The new find_openssh_bignum_offset()
is independent of the gcrypt API. We put this function into
crypt_common.c rather then into gcrypt.c so that openssh.c can also
use it.
Andre Noll [Tue, 28 Aug 2018 19:38:33 +0000 (21:38 +0200)]
crypt: Rename decoding functions.
decode_ssh_key() and decode_key() operate on public and private keys
respectively. This should be obvious from the naming. Also, the
two functions will soon be grouped together in the same file. The
inconsistent naming would even increase the possible confusion,
so it's better to rename the functions now.
Andre Noll [Wed, 29 Aug 2018 10:20:39 +0000 (12:20 +0200)]
gcrypt: Introduce read_openssh_bignum().
This functionality was open-coded twice in apc_get_pubkey() which
reads the two public rsa params. Use the new function instead to
get rid of the duplication. It will be turn out to be useful for
the subsequent work which adds support for RFC4716 keys. Since the
function is specific to libgcrypt, it has to stay in gcrypt.c.
Andre Noll [Tue, 28 Aug 2018 18:50:53 +0000 (20:50 +0200)]
gcrypt: Factor out read_pem_rsa_params().
We'll soon add a similar function which reads bignums from an RFC4716
key file, so it's good to separate this functionality from the rest
of get_private_key().
The patch also introduces struct rsa_params which contains
the six bignums used for RSA private key operations. This way
read_pem_rsa_params() only needs three arguments.
Andre Noll [Tue, 28 Aug 2018 18:18:02 +0000 (20:18 +0200)]
gcrypt: Let read_bignum() return bits, not bytes.
Callers need this to check if the size of the public key is acceptable.
Since sizes of cryptographic keys are generally measured in bits,
its natural to return bits here.
Andre Noll [Mon, 27 Aug 2018 16:32:58 +0000 (18:32 +0200)]
gcrypt: Let decode_key() return blob size through additional argument.
Currently we communicate the blob size through the return value, which
implies an implicit conversion from size_t to int. The new blob_size
argument guarantees type safety and simplifies the logic a bit.
After this patch the return value is only used for the (negative)
error code in the failure case while non-negative return values have
no meaning other than indicating success. Therefore this change also
allows us to convey the key type via the return value. This will turn
out to be handy for supporting RFC4716 keys.
Andre Noll [Sat, 8 Sep 2018 18:31:38 +0000 (20:31 +0200)]
Improve description of para_play.
For one, there is no stop command. Furthermore, make it clear that
the path and the playback position are shown *instead of* the prompt
and the command line.
Andre Noll [Sat, 5 Aug 2017 22:55:50 +0000 (00:55 +0200)]
build: Use .ONESHELL.
This instructs "make" to pass all the lines in the recipe to a
single invocation of the shell. This has no disadvantage and improves
performance slightly. Best out of five "make -j 4" runs:
Before: 0m3.510s
After: 0m3.280s
Speedup: 6.55%
With .ONESHELL given, only the first line of each recipe will be
checked for the special '@' prefix character. This made it necessary
to modify the "Q" macro slightly. Rename it to "SAY" while at it.
The second issue with .ONESHELL is that a failure of any but the
final recipe line will not be noticed by make. To work around this,
we set .SHELLFLAGS to '-ec'. The '-e' flag instructs the shell to
exit at any failure anywhere in the command line.
Andre Noll [Tue, 25 Sep 2018 21:25:42 +0000 (23:25 +0200)]
Remove ->init() of struct receiver.
The previous patch removed the ->init() method of the afh receiver.
Since the afh receiver was the only receiver that defined an init
method, the method can now be removed from struct receiver.
As a consequence, recv_init(), which calls each receiver's init method
has become a no-op and can also be removed.
Andre Noll [Tue, 25 Sep 2018 19:41:20 +0000 (21:41 +0200)]
afh: Constify definition of audio format handlers.
The audio_format_handler structure contains only pointers, and the
->init method of each instance initializes these pointers to constant
values. The ->init() method is thus useless at best, and it prevents
the structures from being declared constant.
This patch removes ->init() of struct audio_format_handler and the
public afh_init() which iterates over all audio formats to call each
->init() method. The audio format handlers are modified to define an
instance of the structure rather than an init function which fills the
fields of the given structure. The structure can be declared constant,
but not static because afh_common.c needs a way to refer to it.
We rely on weak symbols to deal with audio format handlers which are
not compiled in. The codec-independent code in afh_common.c defines
a weak instance of the audio_format_handler structure for each audio
format. The command handlers which are compiled in override the weak
symbol with their own definition.
The afh receiver used to define afh_init() as its (receiver!) init
function, which no longer exists. Since receiver init functions are
optional, we don't need to supply a replacement. However, play.c
calls ->init() of the afh_receiver unconditionally. This call needs
to be removed to avoid a null pointer dereference.
Andre Noll [Fri, 21 Sep 2018 07:29:18 +0000 (09:29 +0200)]
afh: Introduce audio_format_names[].
This removes .name of struct audio_format in favor of an array of
strings. This will allow us to make afl[] a constant array of pointers,
some of which may be NULL to indicate that the audio format was not
compiled in.
This temporarily duplicates the list of audio formats. The second
list will be removed in a subsequent commit.
Andre Noll [Fri, 21 Sep 2018 07:25:46 +0000 (09:25 +0200)]
afh: Get rid of dummy entry at the end of afl[].
The number of audio formats is a compile-time constant. It used to be
the array size of afl[] minus one due to the dummy entry. Without
it, it becomes simply the array size. This patch introduces
NUM_AUDIO_FORMATS as a shortcut for ARRAY_SIZE(afl) and adjusts
next_audio_format() to avoid the access of memory past the end of
the array. With these preparations in place, the dummy entry can
be removed.
Andre Noll [Wed, 5 Sep 2018 10:13:35 +0000 (12:13 +0200)]
blob: Avoid direct access to blob IDs.
Some places of blob.c use read_u32() to get the blob ID from the
osl object while others directly access the mapped memory through
a uint32_t pointer. Besides alignment issues, this causes trouble
on big endian machines. Make sure we always use the primitives from
portable_io.h.
Andre Noll [Wed, 19 Sep 2018 19:08:14 +0000 (21:08 +0200)]
spxdec: Use read_u16() from portable_io.h.
It's about 1% slower on my little endian system, but this is not
reason enough to keep the le_short() macro of spxdec.c. If anything,
the primitives in portable_io.h should be improved.
Andre Noll [Wed, 19 Sep 2018 18:50:01 +0000 (20:50 +0200)]
server: Remove declarations of sender init functions.
The init functions of the three senders have been made static long ago.
Since then vss_init() calls these functions indirectly through the
function pointer of the sender struct. Hence the declarations in
vss.c are no longer needed.
Andre Noll [Wed, 1 Aug 2018 16:29:05 +0000 (18:29 +0200)]
oss: Bail out on unsupported audio formats.
This is better than assuming signed 16 bit little endian. The patch
also improves the error message in oss.c that gets printed when the
audio format could not be set.