Traditionally, many of ALSA library functions are supposed to be
thread-unsafe, and applications are required to take care of thread
safety by themselves. However, people never be careful enough, and
almost all applications fail in this regard.
This patch is an attempt to harden the thread safety in exported PCM
functions in a simplistic way: just wrap some of exported functions
with the pthread mutex of each PCM object. Not all API functions are
wrapped by the mutex since it doesn't make sense. Instead, the
patchset covers only the functions that may be likely called
concurrently. The supposedly thread-safe API functions are marked in
the document.
For achieving the feature, two new fields are added snd_pcm_t when the
option is enabled: thread_safe and lock. The former indicates that
the plugin is thread-safe that doesn't need this workaround and the
latter is the pthread mutex. Currently only hw plugin have
thread_safe=1. So, the most of real-time sensitive apps won't be
influenced by this patchset.
Although the patch covers most of PCM ops, a few snd_pcm_fast_ops are
left without the extra mutex locking: namely, the ones that may have
blocking behavior, i.e. resume, drain, readi, writei, readn and
writen. These are supposed to handle own locking in the callbacks.
Also, if anyone wants to disable this new thread-safe API feature, it
can be still turned off via --disable-thread-safety configure option.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Most of open functions in alsa-lib have the call pattern:
snd_config_update();
return snd_xxx_open(x, snd_config, ...);
This means that the toplevel config gets updated, and passed to a
local open function. Although snd_config_update() itself has a
pthread mutex to be thread safe, the whole procedure above isn't
thread safe. Namely, the global snd_config tree may be deleted and
recreated at any time while the open function is being processed.
This may lead to a data corruption and crash of the program.
For avoiding the corruption, this patch introduces a refcount to
config tree object. A few new helper functions are introduced as
well:
- snd_config_update_ref() does update and take the refcount of the
toplevel tree. The obtained config tree has to be freed via
snd_config_unref() below.
- snd_config_ref() and snd_config_unref() manage the refcount of the
config object. The latter eventually deletes the object when all
references are gone.
Along with these additions, the caller of snd_config_update() and
snd_config global tree in alsa-lib are replaced with the new helpers.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This header isn't used in these files, so stop trying to include it.
This helps builds on systems that don't provide the header (such as
Android/Bionic).
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
When method snd_pcm_set_params sets sample rate to 22050 and latency to 50000
to davinci soc driver method snd_pcm_hw_params_set_buffer_time_near fails
and variable params is already changed in the method so the next method
snd_pcm_hw_params_set_period_time_near fails also.
Signed-off-by: Martin Geier <martin.geier@streamunlimited.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Enable kernel-side functionality by letting user select what sort of
timestamp it desires
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
extend support to link, link_estimated and link_synchronized
timestamp. wall-clock is deprecated
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Like pcm and rawmidi, each object parser needs to accept the hint
component. Now a new local function _snd_conf_generic_id() was
introduced to replace each call of "comment" and "type" field checks.
Also, the two existing identical functions for pcm and rawmidi are
removed and the new function is used commonly.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The size argument is wrong for one of the snprintf() calls in
snd_pcm_chmap_print(), allowing an overflow to happen (the user-provided
buffer may be written data up to 2x its actual size).
Seen in an user report here: http://trac.kodi.tv/ticket/15641
Signed-off-by: Anssi Hannula <anssi.hannula@iki.fi>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Just a couple of quick fixes related to tstamp_type.
- snd_pcm_sw_params(): copy tstamp_type field
- snd_pcm_tstamp_type_name(): fix argument type
- snd_pcm_dump_sw_setup(): fix dumping of tstamp_type field
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch adds big-endian DSD sample format and fixes return value of DSD
formats for snd_pcm_format_little_endian().
Signed-off-by: Jussi Laako <jussi@sonarnerd.net>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Add the new DSD_U32_LE sample format to alsa-lib.
NB include/pcm.h and include/sound/asound.h are updated so a new sync with the
kernel headers is not needed
Signed-off-by: Jurgen Kramer <gtmkramer@xs4all.nl>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Fill the new proto field introduced to sw_params with the current PCM
protocol version. This makes tstamp_type evaluated properly in the
kernel.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
For obtaining / changing the timestamp type, add the corresponding
sw_params accessor API functions together with the public definitions
of timestamp types.
This patch only adds the functions and defines but doesn't bring the
functional changes yet.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
If we try to set the channel map to what it already is, this should
always succeed. E g, speaker-test can do this sometimes.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In particular, this adds definitions and descriptions for G.723 and
DSD types.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Upon an interrupt, it is necessary to abort the wait loops with the EINTR
error code. Introduce snd_*_abort() functions to handle this case.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The memcpy() function in snd_pcm_area_copy() should not be called
with the overlapped areas. Alex discovered - using own LD_PRELOAD checked
for memcpy() input - that the memcpy() is called with src == dst.
For some special plugin combos (rate+softvol+hw for example), the same
areas with same offsets can be asked to be copied (softvol). The collapse
check uses own areas created on heap, causing dst_area == src_area &&
dst_offset == src_offset check bypassed.
Two fixes are in this patch:
- use assert to check the memcpy() input for future triggers
- bypass the snd_pcm_area_copy() call for collapsed identical areas
Reported-by: Alexander Kruppa <akruppa@gmail.com>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Fix the call of _snd_pcm_hw_params() to _snd_pcm_hw_params_internal().
The build passed wrongly because of _snd_pcm_hw_params define.
Having the same function name as the struct name is is simply
confusing...
Also, to be sure, _snd_pcm_hw_params is undef'ed in pcm_local.h, too.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Instead of keeping modified asound.h and asequencer.h, copy the files
in the kernel include/uapi/sound as is, and give some renames for
avoiding conflicts with alsa-lib's definitions.
Some structs, unions and typedefs in asound.h and asequencer.h are
once renamed in local.h before inclusion, then renamed back again.
A bonus by this action is that some local codes don't have to refer to
sndrv_xxx any longer. However, some codes like src/seq/seq.c need
explicit cast because of unavoidable conflicts of struct types.
Another significant change is that now snd_pcm_sw_params.period_event
field is removed. Instead, try to access the last reserved field
internally.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
snd_pcm_write_areas() tries to wait until avail >= avail_min condition
is satisfied. This doesn't work always well when a rate plugin is in
the play.
When a partial data with a smaller size than a period is written, the
rate plugin doesn't transfer the data immediately to the slave PCM,
but kept in an internal buffer and it changes only the hwptr of the
plugin. Thus, the condition "avail < avail_min" is triggered for a
wait check although the underlying slave PCM has enough room. This
results in a call of snd_pcm_wait() which returns immediately after
poll() call, and the snd_pcm_write_areas() loop continues. As a
consequence, it falls into a CPU hog.
This patch fixes that busy loop by introducing a new fast_ops to check
the availability for wait of avail_min. Then a plugin can ask the
slave PCM whether the wait is required (or possible).
A few plugins like multi plugin need a special handling. Otherwise a
generic plugin function can be used.
Reported-by: Trent Piepho <tpiepho@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The init-only plugins do not have own pcm handle, so free the references
to open function immediately after open.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
add new snd_pcm_status_get_audio_htstamp() routine to
query the audio timestamps provided by the kernel.
This change provides applications with better ways
to track elapsed time. Before this patch, applications
would subtract queued samples (delay) from written samples,
resulting in a 1-2 sample error.
Also add snd_pcm_hw_params_supports_audio_wallclock_ts()
to query what the hardware supports.
TODO: check protocol compatibility?
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Follow the new definitions in the kernel side. MONO and others have
been added, and the order of position table was changed again.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Instead of a single channel map, multiple channel maps can be provided
in a form of compound (array) to hw and null plugins. In null
get_chmap, the channel map corresponding to the current channels is
copied from the given channel maps.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Add a config definition "chmap" to override (or enhance) the channel
maps. So far, only a single channel map can be provided, and the
channel count consistency isn't strictly tested at all.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Added a few helper functions between chmap and string.
snd_pcm_chmap_type_name() -- a string of the given chmap type
snd_pcm_chmap_name() -- a string of the given channel position
snd_pcm_chmap_print() -- print channel map on the given buffer
snd_pcm_chmap_from_string() -- get a channel position from string
snd_pcm_parse_string() -- parse the whole channel map from string
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Instead of passing ambiguous integer array, define snd_pcm_chmap_t and
snd_pcm_chmap_query_t so that user can understand more easily which
element is for what.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Added new channel-mapping API functions.
Not all plugins are covered, especially the route, multi and external
plugins don't work yet.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Of course, 24bit samples are not stored in three bits, but bytes.
Signed-off-by: Adrian Knoth <adi@drcomp.erfurt.thur.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The rules engine avoids recalculating rules that do not depend on
any changed parameter, but there is no mechanism to record changed
flags. So when we change a flag, we have to ensure that all rules
depending on that flag are recalculated; the only method to do this
is to force recalculation of all rules.
So far, there have been no kernel drivers with rules depending
on flags, but rules to disable hardware SRCs by setting
SND_PCM_HW_PARAMS_NORESAMPLE are being introduced now.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
This patch adds new API functions, snd_ctl_open_fallback() and
snd_pcm_open_fallback(). These are just like snd_*_open_lconf() but
used to open a fallback PCM/control. The difference is that it replaces
the name string with the given original name, so that hctl and other
upper-layers will recognize it as an alias.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
If we try to print this then it causes a read past the end of the array
in snd_pcm_type_name().
From: Dan Carpenter <error27@gmail.com>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
For capture stream, the delay must be obtained as last, but we need to
update the ring buffer pointers for the avail_update call. So, rearrange
the code a bit and add hwsync call as first.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Add API to disable period wakeups.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@intel.com>
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
- changed logic to get/put blocks
- added mutex locking of the symbol list
- added reference counting (do not free used dl handles)
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Having getenv directly in the function is not a good idea. Allow
configuration of the behaviour change in snd_pcm_param_choose() function
using alsa.conf. Use the environment variable LIBASOUND_COMPAT only
as fallback.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
