Now that we have the necessary infrastructure to memexport and
mempimport a memfd memblock, extend that support higher up in the
chain with pstreams.
A PA endpoint can now _transparently_ send a memfd memblock to the
other end by simply calling pa_pstream_send_memblock() – provided
the block's memfd pool was earlier registered with the pstream.
If the pipe does not support memfd transfers, we fall back to
sending the block's full data instead of just its reference.
** Further details:
A single pstream connection usually transfers blocks from multiple
pools including the server's srbchannel mempool, the client's
audio data mempool, and the server's global core mempool.
If these mempools are memfd-backed, we now require registering
them with the pstream before sending any blocks they cover. This
is done to minimize fd passing overhead and avoid fd leaks.
Moreover, to support all these pools without hard-coding their
number or nature in the Pulse communication protocol itself, a new
REGISTER_MEMFD_SHMID command is introduced. That command can be
sent _anytime_ during the pstream's lifetime and is used for
creating on demand SHM ID to memfd mappings.
Suggested-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
Color global mempools with a special mark. This special marking
is needed for handling memfd-backed pools.
To avoid fd leaks, memfd pools are registered with the connection
pstream to create an ID<->memfd mapping on both PA endpoints.
Such memory regions are then always referenced by their IDs and
never by their fds, and so their fds can be safely closed later.
Unfortunately this scheme cannot work with global pools since the
registration ID<->memfd mechanism needs to happen for each newly
connected client, and thus the need for a more special handling.
That is, for the pool's fd to be always open :-(
Almost all mempools are now created on a per-client basis. The
only exception is the pa_core's mempool which is still shared
between all clients of the system.
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
To transfer memfd-backed blocks without passing their fd every time,
thus minimizing overhead and avoiding fd leaks, a command is sent
with the memfd fd as ancil data very early on.
This command has an ID that uniquely identifies the memfd region.
Further memfd block references are then exclusively done using this
ID.
This commit implements the details of such 'permanent' mappings on
the receiving end, using memimport segments.
Suggested-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
Soon we're going to have three types of memory pools: POSIX shm_open()
pools, memfd memfd_create() ones, and privately malloc()-ed pools.
Thus introduce annotations for the memory types supported and change
pa_mempool_new() into a factory method based on required memory.
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
In future commits, server-wide SHMs will be replaced with per-client
ones that will be dynamically created and freed according to clients
connections open and close.
Meanwhile, current PA design does not guarantee that the per-client
mempool blocks are referenced only by client-specific objects.
Thus reference count the pools and let each memblock inside the pool
itself, or just attached to it, increment the pool's refcount upon
allocation. This way, per-client mempools will only be freed when no
further component in the system holds any references to its blocks.
DiscussionLink: https://goo.gl/qesVMV
Suggested-by: Tanu Kaskinen <tanuk@iki.fi>
Suggested-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
This is needed for building with anonymous unions. A bunch of calls to
fail() that used to mysteriously work need fixing -- fail() is a macro
that takes a printf-style message as an argument. Not passing this
somehow worked with the previous compiler flags, but breaks with
-std=c11.
There is no way to check CPU type in a portable way across ABIs.
Assume if pointers are 64-bit that CPU is capable to perform fast
64-bit operations. Add an extra check to handle x32-ABI.
PulseAudio by default builds with -Wundef. If we add -Werror=undef this
missing define is fatal. By default build log is full of entries like:
In file included from ./pulsecore/core.h:47:0,
from ./pulsecore/module.h:31,
from ./pulsecore/sink-input.h:31,
from pulsecore/sound-file-stream.c:36:
./pulsecore/sample-util.h: In function 'pa_mult_s16_volume':
./pulsecore/sample-util.h:58:5: warning: "__WORDSIZE" is not defined [-Wundef]
#if __WORDSIZE == 64 || ((ULONG_MAX) > (UINT_MAX))
^
(NetBSD-7.99.21 with default GCC 4.8.5)
This change fixes build issues on NetBSD.
This also address a bug reported by Shawn Walker from Oracle (possibly Solaris):
Bug 90880 - builds can fail due to non-portable glibc-specific internal macro usage
sync-playback just had a much longer timeout than it should have, and
extended-test was using the default. We set the expected amount of time,
so the test is more correct (if it takes longer than this, something
probably actually broke).
Pulseaudio fails to build on the Alpha architecture due to a failure
in the volume-test of the test suite. I had reported this to the
Debian bug tracker [1] but the maintainer has asked that I forward the
patch to this mail list. The failure in volume-test occurs because it
is compiled with -ffast-math which implies -ffinite-math-only of which
the gcc manual states that it optimizes for floating-point arithmetic
with the assumption that arguments and results are not NaNs or
+/-infinity, and futher notes that it may result in incorrect output.
On the Alpha platform that is somewhat an understatement as the use of
non-finite floating-point arithmetic with -ffinite-math-only results in
a floating-point exception and the termination of the program.
The volume-test converts volumes into decibels (so a zero volume
becomes a negative infinity) and proceeds to add two volumes (in
decibels), thus does arithmetic with non-finite floating point numbers
despite being compiled with -ffast-math!
I attach a patch that protects against the arithmetic with non-finite
numbers for your consideration. With that patch the test-suite passes
on Alpha.
Cheers
Michael.
[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=798248
...otherwise this code will fail on big-endian architectures.
Cc: Hui Wang <hui.wang@canonical.com>
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
On 32bits OS, this test case fails. The reason is when rewinding to
the middle of a block, some of float parameters in the saved_state
are stored in the memory from FPU registers, and those parameters will
be used for next time to process data with lfe. Here if FPU register
is over 32bits, the storing from FPU register to memory will introduce
some variation, and this small variation will introduce small
variation to the rewinding result.
So adding the tolerant variation for comparing the rewind result, make
this test case can work on both 64bits OS and 32bits OS.
Signed-off-by: Hui Wang <hui.wang@canonical.com>
so far, this test only includes rewind test, it works as below:
let lfe-filter process 2 blocks mono lfe channel audio samples, the
sample format is PA_SAMPLE_S16LE, save the processed data to the temp
buffer, then rewind the lfe-filter back 1 block and 1.5 blocks
respectively, reprocess the audio samples from the rewind position,
then comparing the output data with previously saved data.
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Add a user defined parameter lfe-crossover-freq for the lfe-filter,
to pass this parameter to the lfe-filter, we need to change the
pa_resampler_new() API as well.
Signed-off-by: Hui Wang <hui.wang@canonical.com>
FSF addresses used in PA sources are no longer valid and rpmlint
generates numerous warnings during packaging because of this.
This patch changes all FSF addresses to FSF web page according to
the GPL how-to: https://www.gnu.org/licenses/gpl-howto.en.html
Done automatically by sed-ing through sources.
Because debian does not run with the freebsd libc, but rather uses the
GNU one, it chose to not define __FreeBSD__, but rather __FreeBSD_kernel__.
Use the alternative when the functionality tested is for kernel
features, and keep the __FreeBSD__ one when using freebsd libc
headers.
If this patch is applied, debian could drop all the current patches when
importing 6.0 :)
supresses a warning when compiling with NDEBUG:
pulsecore/aupdate.c: In function 'pa_aupdate_read_end':
pulsecore/aupdate.c:82:14: warning: variable 'n' set but not used [-Wunused-but-set-variable]
unsigned n;
pulsecore/sink-input.c: In function 'pa_sink_input_unlink':
pulsecore/sink-input.c:648:27: warning: variable 'p' set but not used [-Wunused-but-set-variable]
pa_source_output *o, *p = NULL;
pulsecore/sink-input.c: In function 'find_filter_sink_input':
pulsecore/sink-input.c:1523:14: warning: unused variable 'i' [-Wunused-variable]
unsigned i = 0;
pulsecore/sink-input.c: In function 'pa_sink_input_start_move':
pulsecore/sink-input.c:1569:27: warning: variable 'p' set but not used [-Wunused-but-set-variable]
pa_source_output *o, *p = NULL;
CC pulsecore/libpulsecore_5.0_la-sink.lo
pulsecore/sink.c: In function 'pa_sink_unlink':
pulsecore/sink.c:673:24: warning: variable 'j' set but not used [-Wunused-but-set-variable]
pa_sink_input *i, *j = NULL;
pulsecore/source-output.c: In function 'find_filter_source_output':
pulsecore/source-output.c:1179:9: warning: unused variable 'i' [-Wunused-variable]
int i = 0;
CC pulsecore/libpulsecore_5.0_la-source.lo
pulsecore/source.c: In function 'pa_source_unlink':
pulsecore/source.c:616:27: warning: variable 'j' set but not used [-Wunused-but-set-variable]
pa_source_output *o, *j = NULL;
Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
This test broke when PA_MAX_INPUTS_PER_SINK was increased from 32 to 256.
Because we currently don't have time to figure out why, let's just set
NSTREAMS to 20 in the meantime.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
As a third parameter, add the number of samples to read/write in
every iteration. This will help slow CPUs.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
building PA with -O0 leads to test failure in mix-test on i386
issue reported by Felipe, see
http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html
the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according
to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN)
when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing
a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN!
so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh!
the following test code can be used; due to volatile, it will always demonstrate the issue;
without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9):
// snip
static inline float PA_FLOAT32_SWAP(float x) {
union {
float f;
uint32_t u;
} t;
t.f = x;
t.u = bswap_32(t.u);
return t.f;
}
int main() {
unsigned x = 0xbeffbd7f;
volatile float f = PA_FLOAT32_SWAP(*(float *)&x);
printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f);
}
// snip
the problem goes away with optimization when no temporary floating point registers are used
the proposed solution is to avoid passing swapped floating point data in a
float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE()
which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary
float variable is used
also delete PA_FLOAT32_TO_LE()/_BE(), not used
Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
Reported-by: Felipe Sateler <fsateler@debian.org>
Debug and info messages are primarily meant for developers,
rather than end users. Let's save translators' time,
and leave them untranslated.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
to print a pa_usec_t, the format specifier to use is "%" PRIu64
modules/module-combine-sink.c: In function 'update_latency_range':
modules/module-combine-sink.c:750:5: warning: format '%lu' expects argument of type 'long unsigned int', but argument 6 has type 'pa_usec_t' [-Wformat]
modules/module-combine-sink.c:750:5: warning: format '%lu' expects argument of type 'long unsigned int', but argument 7 has type 'pa_usec_t' [-Wformat]
to print a size_t, use %zu
Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
