Commit graph

866 commits

Author SHA1 Message Date
Wim Taymans
15e4e6bcbc pulse-server: remove warning
When we get EPIPE from a client send, pass the error code without
logging a warning. The caller will log an appropriate message instead.

Fixes #2754
2022-10-12 10:17:49 +02:00
Wim Taymans
5b2df0b35c pulse-server: handle unknown format 2022-10-06 16:14:35 +02:00
Wim Taymans
2edfea8b5b module-rtp: add pulse-server modules 2022-10-06 13:18:35 +02:00
Wim Taymans
9b6e504c19 clean up some more array iterations 2022-10-03 09:20:42 +02:00
Wim Taymans
d22feab92a spa: add macro to simplify array iterations some more
uint32_t i;
	for (i = 0; i < SPA_N_ELEMENTS(some_array); i++)
		.. stuff with some_array[i].foo ...

   becomes:

	SPA_FOR_EACH_ELEMENT_VAR(some_array, p)
		.. stuff with p->foo ..
2022-09-30 16:24:26 +02:00
Wim Taymans
323ec0b51b pulse-server: improve maxlength calculations
When no maxlength is given, we use the MAXLENGTH value but we need to
round it DOWN (instead of up) because our buffer is only MAXLENGTH big.
Use CLAMP where we can.
When in capture mode, the maxlength exceeds MAXLENGTH, scale down the
fragsize instead.

Fixes noise in audacious when playing 6 channels sounds. float32 * 6
channels with the maximum buffer size would result in the ringbuffer
being overwritten.
2022-09-26 17:29:17 +02:00
Wim Taymans
e0d1be32fd pulse-server: don't read more than requested size
When we underrun, don't try to read more than what was requested
even if there is more in the ringbuffer.
2022-09-26 13:30:39 +02:00
Demi Marie Obenour
671a7102ff Fix some badly-behaved macros
Some macros evaluated their arguments more than once when it was not
needed, or were missing parentheses.
2022-09-23 11:39:33 -04:00
Demi Marie Obenour
bb4f274ae0 Make all fopen() calls use O_CLOEXEC
by adding "e" to the mode strings.
2022-09-23 15:19:01 +00:00
Wim Taymans
e0614246f5 pulse-server: set fragsize as latency
Reducing the latency is just papering over the issue in #2702. The
real fix is to limit the blocksize to the fragsize like what is done
in 00a234daf2

Reducing the latency then also causes regressions like #2715 so don't
do that anymore.

Fixes #2715
2022-09-22 11:58:14 +02:00
Wim Taymans
47f2b72088 pulse-server: don't overwrite previous towrite value
And actually take the MAX_BLOCK into account when writing samples to a
client.
2022-09-21 12:00:36 +02:00
Wim Taymans
00a234daf2 pulse-server: limit capture blocks to fragsize
Don't send larger chunks than the fragsize. Some clients don't like
this. Also checked with #2418.

Fixes #2711
2022-09-21 11:34:13 +02:00
Wim Taymans
bd584ca8c0 pulse-server: ignore adjust-latency for capture
Don't update the fragsize, it breaks #2418 again.
2022-09-21 11:34:13 +02:00
Wim Taymans
0ce95e69fd pulse-server: adjust_latency should update the fragsize
Rename MAX_FRAGSIZE to MAX_BLOCK because it does not really limit the
fragsize, just the amount of data we can send in one block.

See #2711
2022-09-20 13:24:30 +02:00
Wim Taymans
d1cd199f16 pulse-server: set record latency as fragsize / 2
The record latency is set to fragsize / 2 by pulseaudio, so do the
same in pipewire.

Fixes #2702
2022-09-19 14:45:11 +02:00
Wim Taymans
56b03cb3a8 pulse-server: update requested for relative seek only
PulseAudio only updates the requested amount of bytes for relative
seeks, do the same.

See #2680
2022-09-10 16:06:45 +02:00
Barnabás Pőcze
27211856eb pulse-server: Handle overlong writes from clients
When a client writes more then requested, let the requested field go
negative so that it is taken into account the next time we ask for more
data.

Also the requested field follows the difference in the write pointer
caused by seeks.

See #2626
Fixes #2674
2022-09-08 09:14:19 +02:00
Wim Taymans
d5bc6aa2cb Revert "pulse-server: Handle clients that send more than requested"
This reverts commit a79b5c86ea.
2022-09-08 09:06:02 +02:00
Wim Taymans
a79b5c86ea pulse-server: Handle clients that send more than requested
A client can sometimes send more data than we requested. PulseAudio
keeps the extra data around, it just asks for more data when it consumed
some of it.

PipeWire however always tries to keep tlength worth of data, as
specified in the PulseAudio docs... Keep track of how much extra data
has been sent and keep this around as well. Make sure we flush this
extra data as well.

Fixes #2626
2022-09-07 13:07:28 +02:00
Wim Taymans
50f76c8730 pulse-server: implement TRIGGER and PREBUF methods
Start and stop prebuf.
2022-09-07 13:02:55 +02:00
Wim Taymans
305f2104ee pulse-server: rework buffer attributes
Move the latency fraction calculation to fix_ functions so that the
new latency rate can be used when creating the streams.
Actually set the requested record attributes on the stream instead
of modifying the defaults.

See #2671
2022-09-06 11:53:51 +02:00
Wim Taymans
1a915c2ced pulse-server: set record attributes more like PulseAudio
PulseAudio assigns half of the fragsize to the source latency.
PulseAudio sends in chunks up to a fixed limit.

Checked against #2418
2022-09-05 18:13:12 +02:00
Wim Taymans
ad06db91ee Revert "pulse-server: send chunks with half the fragsize"
This reverts commit e10305851d.

If causes some regressions: see #2671
2022-09-05 16:19:42 +02:00
Wim Taymans
83e83608c8 Revert "pulse-server: tweak the fragsize/minreq handling"
This reverts commit ad214dde2e.
2022-09-05 16:18:54 +02:00
Wim Taymans
009ab9681d Revert "pulse-server: simplify latency handling"
This reverts commit 9c34a923db.
2022-09-05 16:18:45 +02:00
Wim Taymans
72ce2f815b Revert "pulse-server: enforce minfrag after latency adjust"
This reverts commit b70f32a259.
2022-09-05 16:18:34 +02:00
Wim Taymans
b70f32a259 pulse-server: enforce minfrag after latency adjust 2022-09-03 13:07:09 +02:00
Wim Taymans
9c34a923db pulse-server: simplify latency handling
We don't need to use the minreq field. The fragsize is only updated when
adjust_latency is used.
2022-09-01 16:46:18 +02:00
Wim Taymans
ad214dde2e pulse-server: tweak the fragsize/minreq handling
Only send 2/3 of the fragsize
2022-09-01 16:23:56 +02:00
Wim Taymans
e10305851d pulse-server: send chunks with half the fragsize
PulseAudio configures half of the fragsize as the source latency. It
also sends chunks as soon as they become available.

This means that we also need to configure the source with half of
the fragsize latency and send in chunks of fragsize/2. Keep this in
the unused (for record) minreq field.
2022-09-01 15:46:17 +02:00
Pauli Virtanen
6d4d77802a pulse-server: give Manager permission for Flatpak apps with devices=all
Some Flatpak apps want to manage sound fully (change default outputs,
move streams from other applications).  Real Pulseaudio always grants
full permissions, but pipewire-pulse doesn't, which breaks some
applications, e.g. Zoom.

Work around this by granting the manager permission if the application
also has devices=all access.

To do things properly, this probably should use the Portal media roles,
but this would need further work elsewhere.
2022-08-01 19:36:00 +00:00
Pauli Virtanen
0e12242a1b modules: set pipewire.access.portal.app_id in access & pulse-server
When detecting a access=flatpak application, set also
pipewire.access.portal.app_id, if determined
2022-08-01 19:36:00 +00:00
Pauli Virtanen
3730e54484 modules: support getting app_id in pw_check_flatpak
Support getting the Flatpak application ID from the .flatpak-info file,
similarly as what xdg-desktop-portal does.
2022-08-01 19:36:00 +00:00
Pauli Virtanen
5e5339ce6c modules: refactor check_flatpak to a common place
Move check_flatpak used by module-access and module-protocol-pulse
to a common flatpak-utils.h
2022-08-01 19:36:00 +00:00
Wim Taymans
35082e4f4b pulse-server: add the delay
Actually add the delay to the buffered amount of time instead of
overwriting.
2022-07-13 10:49:26 +02:00
Wim Taymans
855d17ac1b pulse-server: add some debugging 2022-07-13 10:42:00 +02:00
Pauli Virtanen
167537dffc pulse-server: fix warning
The rest of the fields are zeroed later in reply_create_record_stream,
but static analysis doesn't know that.
2022-07-06 14:36:22 +00:00
Pauli Virtanen
a46d4aedd7 pulse-server: work around race condition in SET_DEFAULT/MOVE_*
The default metadata update may be delayed for some time after
SET_DEFAULT, since the session manager has to work on it.

MOVE_* commands need to know what the client thinks is the current
default sink, so that they can decide between "move to default" and
"move to specific device".

To avoid race condition here, stash the default value set by the client
and use it in the MOVE_* commands, until the metadata update arrives.
2022-07-04 15:57:31 +00:00
Adolfo Rodrigues
6188e4d074 pulse-server: tweak maxlength some more
Make the buffer size at least 4 times as large as the fragment size to avoid
xruns.
2022-07-04 15:56:16 +00:00
Wim Taymans
8b649c4510 pulse-server: optimize property allocations
Only allocate a new properties object when we actually need to
change something.
2022-07-02 11:27:54 +02:00
Davide Beatrici
a31989b46d pulse-server: add card properties to sinks and sources
To match the original PulseAudio server's behavior.
2022-07-01 19:54:44 +00:00
Davide Beatrici
e0731f0ffa pulse-server: map "device.sysfs.path" to "sysfs.path" 2022-06-30 08:41:30 +02:00
Barnabás Pőcze
97a275e71e pulse-server: update quirks before querying the name
`client_update_quirks()` may modify `client::props` depending
on the "pulse.rules" section defined in the configuration file,
which means that it can change the value corresponding to the
"application.name" key. Therefore, `client::name` needs to
be queried after that.
2022-06-20 17:30:05 +00:00
Barnabás Pőcze
0f45d784d8 pulse-server: update client::name on UPDATE_CLIENT_PROPLIST command
`client::name` points to a string that is owned by `client::props`,
so when the property list is updated, it needs to be refreshed as well.
Otherwise, various use-after-frees can be triggered, for example:

==1471586==ERROR: AddressSanitizer: heap-use-after-free on address 0x60200007e7d0 at pc 0x7f14390755d0 bp 0x7ffe23edee30 sp 0x7ffe23ede5a8
READ of size 3 at 0x60200007e7d0 thread T0
    #0 0x7f14390755cf in printf_common /usr/src/debug/gcc/libsanitizer/sanitizer_common/sanitizer_common_interceptors_format.inc:553
    #1 0x7f1439077215 in __interceptor_vsnprintf /usr/src/debug/gcc/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:1665
    #2 0x7f1434ead47d in spa_vscnprintf ../spa/include/spa/utils/string.h:239
    #3 0x7f1434eae2ae in impl_log_logtv ../spa/plugins/support/logger.c:138
    #4 0x7f14385cacc7 in pw_log_logt ../src/pipewire/log.c:135
    #5 0x7f1433aef8e9 in do_set_profile ../src/modules/module-protocol-pulse/pulse-server.c:4656
    #6 0x7f1433b0af4d in handle_packet ../src/modules/module-protocol-pulse/server.c:109
    #7 0x7f1433b0e747 in do_read ../src/modules/module-protocol-pulse/server.c:276
    #8 0x7f1433b0eb04 in on_client_data ../src/modules/module-protocol-pulse/server.c:306
    #9 0x7f1434ec56a0 in source_io_func ../spa/plugins/support/loop.c:442
    #10 0x7f1434ec4a21 in loop_iterate ../spa/plugins/support/loop.c:430
    #11 0x7f14385dc23d in pw_main_loop_run ../src/pipewire/main-loop.c:148
    #12 0x55b065d73722 in main ../src/daemon/pipewire.c:131
    #13 0x7f143742928f  (/usr/lib/libc.so.6+0x2928f)
    #14 0x7f1437429349 in __libc_start_main (/usr/lib/libc.so.6+0x29349)
    #15 0x55b065d722a4 in _start (./src/daemon/pipewire-pulse+0x42a4)

0x60200007e7d0 is located 0 bytes inside of 16-byte region [0x60200007e7d0,0x60200007e7e0)
freed by thread T0 here:
    #0 0x7f14390be672 in __interceptor_free /usr/src/debug/gcc/libsanitizer/asan/asan_malloc_linux.cpp:52
    #1 0x7f14386a775a in do_replace ../src/pipewire/properties.c:414
    #2 0x7f14386a785e in pw_properties_set ../src/pipewire/properties.c:441
    #3 0x7f14386a658b in pw_properties_update ../src/pipewire/properties.c:309
    #4 0x7f1433adb055 in do_update_proplist ../src/modules/module-protocol-pulse/pulse-server.c:3246
    #5 0x7f1433b0af4d in handle_packet ../src/modules/module-protocol-pulse/server.c:109
    #6 0x7f1433b0e747 in do_read ../src/modules/module-protocol-pulse/server.c:276
    #7 0x7f1433b0eb04 in on_client_data ../src/modules/module-protocol-pulse/server.c:306
    #8 0x7f1434ec56a0 in source_io_func ../spa/plugins/support/loop.c:442
    #9 0x7f1434ec4a21 in loop_iterate ../spa/plugins/support/loop.c:430
    #10 0x7f14385dc23d in pw_main_loop_run ../src/pipewire/main-loop.c:148
    #11 0x55b065d73722 in main ../src/daemon/pipewire.c:131
    #12 0x7f143742928f  (/usr/lib/libc.so.6+0x2928f)

previously allocated by thread T0 here:
    #0 0x7f1439072faa in __interceptor_strdup /usr/src/debug/gcc/libsanitizer/asan/asan_interceptors.cpp:439
    #1 0x7f14386a6fe2 in do_replace ../src/pipewire/properties.c:394
    #2 0x7f14386a785e in pw_properties_set ../src/pipewire/properties.c:441
    #3 0x7f1433a6c52d in read_props ../src/modules/module-protocol-pulse/message.c:147
    #4 0x7f1433a6f467 in message_get ../src/modules/module-protocol-pulse/message.c:359
    #5 0x7f1433ab3191 in do_set_client_name ../src/modules/module-protocol-pulse/pulse-server.c:1030
    #6 0x7f1433b0af4d in handle_packet ../src/modules/module-protocol-pulse/server.c:109
    #7 0x7f1433b0e747 in do_read ../src/modules/module-protocol-pulse/server.c:276
    #8 0x7f1433b0eb04 in on_client_data ../src/modules/module-protocol-pulse/server.c:306
    #9 0x7f1434ec56a0 in source_io_func ../spa/plugins/support/loop.c:442
    #10 0x7f1434ec4a21 in loop_iterate ../spa/plugins/support/loop.c:430
    #11 0x7f14385dc23d in pw_main_loop_run ../src/pipewire/main-loop.c:148
    #12 0x55b065d73722 in main ../src/daemon/pipewire.c:131
    #13 0x7f143742928f  (/usr/lib/libc.so.6+0x2928f)
2022-06-20 17:30:05 +00:00
Wim Taymans
ff05ebada6 pulse-server: tweak record attributes
Add some 'before' debug info of the attributes.
First clamp the fragsize to the maxlength and then clamp to the
minfrag or else we could bypass the configured minfrag by setting a
lower maxlength.
Scale the maxlength to at least twice the fragsize so that we can
buffer enough data in the ringbuffer to serve fragsize bytes.

See #2447
2022-06-20 11:42:59 +02:00
Lucas Holt
6a15a02ec2 Add support for MidnightBSD
Fix build issue

Fix build issue
2022-06-19 18:22:47 +00:00
Barnabás Pőcze
ae34b68e3a pulse-server: message: improve resizing logic
As Coverity points out, `ensure_size()` is not fully correct.

Let us assume that the message already has some allocated storage,
and the `realloc()` call in the next `ensure_size()` invocation fails.
In that case `message::data` is freed, but the pointer is left behind.
If another `ensure_size()` call is made, then `realloc()` will be called
(since the previous call left `message::allocated` as zero), but the
first argument of the `realloc()` call will be a dangling pointer.

In order to avoid the above, first of all, clear `message::data` after
a failed `realloc()` call, and immediately return `-ENOMEM` if
`message::length` is greater than `message::allocated` since
that signals if the message has even run into an out-of-memory
situation.
2022-06-16 18:38:30 +00:00
Barnabás Pőcze
080a97c0bb pulse-server: message: store pointer to impl directly
Store a pointer to the owner `impl` object instead of
the (embedded) `stat` object. This way `message_free()`
can be simplified since the owner `impl` does not need
to be passed explicitly.
2022-06-16 18:38:30 +00:00
Barnabás Pőcze
b843b0ab29 pulse-server: message: remove unnecessary forward declaration
`struct client` is never referenced in "message.h",
so remove the forward declaration for it.
2022-06-16 18:38:30 +00:00
Wim Taymans
68bafcf166 pulse-server: Don't send more than fragsize chunks
Send exactly fragsize chunks when capturing.

See #2418 and #2434
2022-06-14 11:20:18 +02:00