In the default configuration, PulseAudio's rlimit-rttime is set to
1000000 (100%), which is higher than what RealtimeKit requires from
its clients (200000, 20%).
Make an attempt to still get realtime scheduling by clamping the
current RLIMIT_RTTIME to what RealtimeKit accepts. Warn about doing
this.
Any code that runs inside the init() callback sees an invalid module
index. Sometimes init() does things that cause hooks to be fired. This
means that any code that uses hooks may see an invalid module index.
Fix this by assigning the module index before init() is called.
There are no known issues in the upstream code base where an invalid
module index would be used, but an out-of-tree module
(module-murphy-ivi) had a problem with this.
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=63923
pa_write() knows two types of operation:
calling send() and calling write()
there is a flag (a pointer to an int) passed to pa_write()
which can remember which write type was successful
if the pointer is NULL or the int is 0, send() is tried first,
with a fallback to write() if send() resulted in ENOTSOCK
pa_fdsem_post() calls pa_write() with a NULL pointer;
unfortunately (at least with HAVE_SYS_EVENTFD_H #define'd) send()
always fails here and write() is called -- causing an extra syscall
quite frequently
strace:
send(17, "\1\0\0\0\0\0\0\0", 8, MSG_NOSIGNAL) = -1 ENOTSOCK (Socket operation on non-socket)
write(17, "\1\0\0\0\0\0\0\0", 8) = 8
the patch adds a write_type field to pa_fdsem to the successful
pa_write() type can be remembered and unnecessary send() calls are
avoided
Signed-off-by: Peter Meerwald <p.meerwald@bct-electronic.com>
The old code accepted any word that started with "y", "Y",
"n", "N", "t", "T", "f" or "F". Fix this by having
a whitelist of full strings instead of checking just the
first letter.
Port creation is now slightly different. It is now similar to how
other objects are created (e.g. sinks/sources/cards).
This should become more useful in the future when we move more stuff to
the ports.
Functionally nothing has changed.
This fixes a later assertion failure in module-stream-restore.
Buglink: https://bugs.launchpad.net/bugs/896602
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
It appears that, libltdl will find the .la file in the builddir and
figure out where the real .so is.
This also requires .ifexists to be fixed up to correspondingly search in
<dlsearchpath>/.libs.
The old assumption seemed to be that if a sink or source has the
DYNAMIC_LATENCY flag set, it can never change, so the fixed latency
will always be zero. This assumption doesn't hold with filter sinks
and sources that are moved around.
This fixes a crash with two module-virtual-sink instances on top of
each other, when the bottom one is moved from a sink without dynamic
latency to a sink with dynamic latency. What happened was that first
the bottom virtual sink "updated" (due to this bug nothing was
actually updated) its fixed latency to match the master sink (zero
fixed latency), and then the top virtual sink updated its fixed
latency to match the master sink. The master sink was the bottom
virtual sink, whose fixed latency should have been set to zero, but it
was not, so the pa_sink_set_fixed_latency_within_thread() failed in
the assertion "latency == 0".
An example: let's say that there's an alsa sink and two filter sinks
on top of each other:
alsa-sink <- filter1 <- filter2
With the old code, if filter1 gets moved to another sink, and the
new sink doesn't have the LATENCY and DYNAMIC_LATENCY flags set
(unlike alsa-sink), filter1's flags are updated fine in the moving()
callback, but filter2 is not notified at all about the flag changes.
With this patch, the flag changes are propagated to filter2 too.
Forcing the shm file to be read-only makes shm_unlink() fail on OS X.
Thanks to Albert Zeyer for reporting the bug and investigating the
root cause.
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=62988
During a stream, most packets sent are either memblocks (with SHM info),
or requests for more data. These are only slightly bigger than the
header.
This patch makes it possible to write these packages in one write
instead of two: a memcpy of just a few bytes is worth saving extra
syscalls for write and poll.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
I don't know if it matters a lot, but most certainly it must be
the new channel that's supposed to be made low-delay, not the existing
listening socket, right?
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
the specialized code path just duplicate samples, so are only
applicable if the volume in map_table is == 1.0 (or == 0x10000);
don't use them for volumes >= 1.0
compare the integer version of the volume stored in map_table;
comparing floats is ugly (als leads to compiler warnings)
Signed-off-by: Peter Meerwald <p.meerwald@bct-electronic.com>
This consumes less power, has low (no?) perceivable difference, and
allows the default configuration to work out of the box on low-end
systems (such as netbooks).
This reverts commit a9c3f2fb0f.
It has been recently agreed that ports should somehow have some physical
meaning, leading to the port merge in module-bluetooth-device.
With this assumption in mind, it is very unlikely that a card would
add or remove ports dynamically. Therefore, the core can be simplified
by removing the support for this.
The revert affects the code added to module-card-restore in commit
a1a0ad1af2, which can now be partially
removed.
Conflicts:
src/pulsecore/card.c
src/pulsecore/core.h
If minreq is not explicitly specified, it was always initialized to
20 ms (DEFAULT_PROCESS_MSEC). However when the total latency is not
much higher than 20 ms, this is way too high. Instead use
tlength/4 as a measure: this will give a decent sink_usec in all
modes (both traditional, adjust latency and early request modes).
This greatly improves PulseAudio's ability to ask for data in time
in low-latency scenarios.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Tlength should never be set higher than maxlength. While this is
corrected by memblockq later, we still need a correct tlength for
the subsequent calculations.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
This reverts commit 6733caf114.
Apparently, the EOF bit gets set only after there has been an attempt
to read more data than the file contains, so just reading the last
byte isn't sufficient.
fgets() returns NULL in case there's an error or f is at EOF. The
while condition just checked that f is not at EOF, therefore an error
must have happened.
Previously, a drain request was acknowledged up to two hw buffers
too late, causing unnecessary delays.
This implements a new chain of events called process_underrun
which triggers exactly when the sink input has finished playing,
so the drain can be acknowledged quicker.
It could later be improved to give better underrun reporting to
clients too.
Tested-by: Dmitri Paduchikh <dpaduchikh@gmail.com>
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Some cards are capable to announce if a specific profile is available or
not, effectively predicting whether a profile switch would fail or would
likely succeed. This can for example be useful for a UI that would gray
out any unavailable profile.
In addition, this information can be useful for internal modules
implementing automatic profile-switching policies, such as
module-switch-on-port-available or module-bluetooth-policy.
In particular, this information is essential when a port is associated
to multiple card profiles and therefore the port availability flag does
not provide enough information. The port "bluetooth-output" falls into
this category, for example, since it doesn't distinguish HSP/HFP from
A2DP.
Generalize the availability flag in order to be used beyond the scope of
ports.
However, pa_port_availability_t is left unchanged to avoid modifying the
protocol and the client API. This should be replaced by pa_available_t
after a validation phase of this new generic enum type.
Unloading modules in the reverse order is the "more logical" thing
to do, and speeds up shutdown somewhat, e g by not loading
module-null-sink at shutdown.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
The previous patch removed module-gconf's dependency on the userdata
pointer of the free callback, and that was the only place where the
userdata pointer of pa_free2_cb_t was used, so now there's no need for
pa_free2_cb_t in pa_hashmap_free(). Using pa_free_cb_t instead allows
removing a significant amount of repetitive code.
