Currently shared memory pointer is initialized to 0 and set to -1
in some, but not in all error paths.
In cleanup path of open the shm pointer is only compared to be non-NULL
before dereferencing it which leads to SEGFAULT in case it was set to -1.
This patch initializes pointer to -1 to have a unique identification
for invalid pointer and also checks for pointer being not -1 on
access in cleanup path.
Signed-off-by: Vanitha Channaiah <vanitha.channaiah@in.bosch.com>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
This commit allows to define private alsa-lib's configuration. When
the configuration is present, the device values ("PlaybackCTL",
"CaptureCTL", "PlaybackMixer", "CaptureMixer", "CapturePCM")
are prefixed with '_ucmHEXA.' string where HEXA is replaced by the
unique hexadecimal number identifying the opened ucm manager handle.
Syntax 4
LibraryConfig.a_label.SubstiConfig {
# substituted library configuration like:
usr_share_dir "${ConfLibDir}"
}
LibraryConfig.b_label.Config {
# non-substituted library configuration like:
usr_share_dir "/usr/share/alsa"
}
The File counterparts:
LibraryConfig.c_label.SubstiFile "/some/path"
LibraryConfig.d_label.File "/some/path"
Note that for files the contents is substituted on the request,
but the file name is always substituted (useful for ${ConfDir} etc.).
The private configuration is not saved or preserved. It's life time
belongs to the opened ucm manager handle.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The appl_ptr difference must be computed before
the appl_ptr in the status structure is updated.
Fixes: da5b70d3f ("pcm: plugin - fix status code for capture")
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The function is not used externally (it's not exported in headers).
Make it private.
Use snd_config_get_ctl_iface() for the interface value parsing.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
It seems that the commit "pcm: ioplug: Transfer all available data"
introduced new regressions (wrong memory access). The second issue
is that the avail_update in ioplug does not move appl_ptr nor hw_ptr,
so it's possible that the transfers may be repetitive.
This patch moves the transfer calls to mmap_begin callback where it
should be. The pointer wraps are handled by design now.
Fixes: 1714332719 ("pcm: ioplug: Transfer all available data")
BugLink: https://github.com/alsa-project/alsa-lib/pull/103
Tested-by: Andreas Pape <apape@de.adit-jv.com>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
This one also has the same problem as others; the appl_ptr and hw_ptr
of share plugin aren't updated in snd_pcm_status() call. Fix it.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Just like the previous fix for ioplug, the null plugin also misses the
appl_ptr and hw_ptr updates for snd_pcm_status(). Fix it as well.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The snd_pcm_status() of the ioplug doesn't return the current
positions of hw_ptr and appl_ptr as advertised. Fix it by copying the
current values stored in the plugin data.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
A snd_pcm_status() call for the direct plugins receives the status
from the slave PCM, but this doesn't contain a valid appl_ptr, since
the slave PCM for the direct plugins is in a free-wheel mode, hence
the appl_ptr is always zero. This result in the inconsistent
status->appl_ptr and pcm->appl.ptr, hitting the recently introduced
assert() call.
Fix it by transferring the plugin's appl_ptr to the upper caller.
BugLink: https://bugzilla.opensuse.org/show_bug.cgi?id=1181194
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Unfortunately, we cannot use status->avail from slave, because this value
does not wrap to the buffer size and it may even overflow slave boundary
(endless run). We can use only hw_ptr from slave.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The direct plugins (dmix & etc.) sets own initial
hw_ptr and appl_ptr. Use this initial settings
to export correct values in snd_pcm_status().
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The recent updates do not take in account the possible
calls for the capture stream. Fix the avail and delay
inconsistencies (and assert).
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
This reverts commit aba87e5098.
The commit does not look good. The plugins must handle
the delay value correctly.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Do hw_ptr sync only once after the status call. This avoids
double update.
Also, the application pointer must not change when
the status is called. Add assert() call for this condition.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
No functional changes - move the code to snd_pcm_plugin_sync_hw_ptr()
and put the mmap capture updates to separate function for readability.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
It's no reason to sync the avail/delay fields using the mirrored
buffer pointers. The slave information must be valid.
The original report probably tries to fix something for
the specific plugin. Revert all changes.
Fixes: afe6ff3b33 ("pcm: plugin status - fix the return value (regression)")
Fixes: 4f90392f07 ("pcm: fix the snd_pcm_plugin_status() avail and delay fields")
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The snd_pcm_plugin_avail_update() error code in snd_pcm_plugin_status()
should not be reported to the caller. The state errors can be determined
using the state member in the status structure.
Fixes: 4f90392f07 ("pcm: fix the snd_pcm_plugin_status() avail and delay fields")
BugLink: https://github.com/alsa-project/alsa-lib/issues/107
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The __snd_pcm_state() must return a valid state, not an error value
when the plugin callback is not defined. Use the first state
SND_PCM_STATE_OPEN - the other functions will return the error
code depending on this state.
Link: https://lore.kernel.org/alsa-devel/20201226213547.175071-10-alexhenrie24@gmail.com/
Reported-by: Alex Henrie <alexhenrie24@gmail.com>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Set the status tstamp field so that it can be accessed with
snd_pcm_status_get_htstamp().
Signed-off-by: Jonas Holmberg <jonashg@axis.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The syscalls may return EINTR when a signal is handled.
Implement safe_write() function which does simple write
retry.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The rewind or forward may cause the stream pointer change. Although
this patch does not fix the real meter update issue, it breaks
the possible big loops when the stream pointers are desynced with
the meters. It does not make sense to copy more samples than the
pcm buffer contains.
Link: https://lore.kernel.org/alsa-devel/f56d6a67-014a-e562-c253-830c0ec03717@ivitera.com/
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
The avail and delay fields in the returned status structure does not
reflect the actual hw_ptr/appl_ptr. This change correct this.
TODO: Unfortunately, the delay might contain also information about
extra hardware / buffering delay which is hidden with this change.
Link: https://lore.kernel.org/alsa-devel/d9c1f37e-5c8d-f289-270e-c6cda7a56ce3@axis.com/
Reported-by: Jonas Holmberg <jonashg@axis.com>
Tested-by: Jonas Holmberg <jonashg@axis.com>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
